Search on the Site
By Josh Lincoln
Posted 14 January 2015, 9.52am est
Searching for "Minster":
Based in ‘Minster’
Kidderminster Hospital Radio
Kidderminster Acute Hospital
WCR Community Radio
Warminster, Westbury and Frome
The Kidderminster Chronicle
Pulman's Weekly News & View from Axminster
Kidderminster & Stourport
Which hell of that is based in Minster? I asked you to see about such abominable fetching last year in the old place, James.
See - what if I start searching for "Q", huh??? I'll get a comprehensive Oxford dictionary then?
I say what - you may add spelling suggestions to the search interface, but please try applying some wiser input processing: like as a part/parts for the main section and as WHOLE WORDS for certain others - I mean including spelling variants to that, like at least plural-singular or such.
Thanks for the feedback. The beauty of searching is that the system has no idea what you have entered. We could check whether it is a town name - but that gives issues for, for example, Hull - which is actually "Kingston-upon-Hull" and therefore wouldn't work.
One-letter searches, like Q, do return data at present. Media UK would refuse a one-letter search. What I should probably do there is, for two and one letter searches, only do a whole word search. There is a good use-case for searching for "Q" - it is a music magazine - but it will get lost in the noise of that type of search.
Spelling variants are not trivial. I know how I would like to code those, but for now if you mistype there is no auto-correct (No "did you mean"). You can use tags for the purpose of adding commonly-used spelling errors. As such, you'll note that a search for "westsound" will also find West Sound.
And as of now, a one or two letter search - like a search for Q - actually does an exact match instead of a 'contains' search.
...but that gives issues for, for example, Hull - which is actually "Kingston-upon-Hull" and therefore wouldn't work.
Unless you have a special database for that purpose. "Hull" is a common name.
And anyway, you could implement a sort of separate search: like "search place" or "search media"; not just piling it all up, as it is now.
One search, one search box, no fiddly drop-downs. That's something I learnt from Media UK v4, which had such a thing: nobody used it.
James, your search seems to get confused by a "technical spacing" occurring when copying things with a triple click then just pasting it into the search box.
Well, in Chrome it'll look like "Preston FM " (no idea if the "spacing"'ll be visible nor how in cases), today in Firefox I couldn't even see such when copypasting (I have a media.info search in my search-box) - it looked normal/looks normal usually and doesn't return any trouble. But now it - worked with that search-box - then your search found--it wasn't "nothing found": some news was there - but no station; then I pasted the same clipboard into the page box directly - the spacing was seen, I removed it, station got found.
Well, most search tools on the web I know of usually lose such spacing or whatever it is automatically. Try testing it yourself, will you?
Thanks for asking so nicely, Josh - but this is now fixed. :)
You might rename it to "Search is insufficient":)
Look, I thought it'd search for ANYTHING rendered as text in respective articles, apart that usually I get some BS about "something IN Chatanooga-choochoo" (while there's no Chatanooga-choochoos on earth). See, now I assumed I'd easily get that "Quay FM" searching for "Alderney".
And yes, both the article and its preview DO HAVE "Alderney" as text etc.
But NO - "NOTHING FOUND". WHY? Where are those CHATANOOGA-CHOOCHOOS when they're needed???
The search function didn't go through media title descriptions.
But I can see how that might be useful; so as of now, it does.
Happy New Year!
I noticed the search engine doesn't search for users registered on the site. It might be useful to have such a search - within or separately.
I've been working on the search today.
Similarly, a UK user for "Virgin Radio" will be offered both Absolute and Virgin, and strangely Virgin second in the list. They should, of course, get Virgin at the top.
The new search code doesn't do this, but is collecting data to enable me to turn this on in the next week or so. A by-product of this is that we've lost station logos and pretty formatting; but that'll come back shortly.
A result doubled again (in one and same section ie) - in a search for
AIR 107.2 (the station's article, same link that is).
As it was stated when creating this thread, it's a mess. Try searching for "CSR", for examle.
"People": a Caesar. Lots of Cats and other animals... Most outrageous I find this "Based in" section.
While for others, we just need to clarify if we want EXACT terms, this one needs amendment. Say, implement a subroutine with a database of places, and if the term doesn't match any, no results will be in there. The database may be either on site, cloud/shared/whatever, or combined.
OR - for this particular section - there'll be a separate search procedure in there: searches ONLY in postal addresses AND editorial areas. In theory, it might be combined with the above proposed using some database(s).
First, for completeness, I should mention that the above geo-aware search is now complete and live. A search for "Smooth" in Australia gives Smooth FM in Sydney, while a search for "Smooth" in the UK gives Smooth London, as the recommended link.
"Based in" - yes, I do have a database of place names. Here's my problem, though - Viking FM is based in Hull, so a search for Hull should find Viking FM. It does. However, the actual name of the place is "Kingston upon Hull", and not just Hull. So, if I implemented code that checked whether "Hull" was a place name and if so, return it... it wouldn't find anything, since Hull isn't actually in that database.
Postal areas and editorial areas are already treated as exact matches, while names are treated as 'sounds like' matches, and so return more results.
''Part of name'' will do it, or the database might get certain adjustments. I don't know how it's organised now, let's say any given place would have a CODE, then this CODED place will have one or more literal variants: say, "Hull", "Kingston-upon-Hul" and "Kingston upon Hull" for Hull (coded), then variants for places in Wales (English and Welsh), Scotland and Ireland (NI included). What do you say?
"Sounds like" is funny:)
"Part of name" is exactly how it works now.
Places already have a code - Hull is ChIJrWABfI8g1kcR_zDxgp1-RG0
Now, this is a good test case, because I did assume that "Hull" wouldn't appear. That's an assumption too far, it seems; since both "Hull" and "Kingston upon Hull" are both ChIJrWABfI8g1kcR_zDxgp1-RG0 - so a search for one should find the other. The only problem is that I don't have those IDs across the directory - only newer items will have that attached to it. It's a big job to go back and add the IDs for others.
"Based in Bute" doesn't return Bute Island Radio anyway. The station is based in the island, isn't it?
Login or register to comment
It only takes a second with your Google or Facebook account.
- follow us on @minfodiscuss