I am using azure search in my bot application.
In this if we give input with spelling mistake, for small words like trvel => travel we are getting response properly.
But if i enter "travelexpense" for this i am not getting any result.
Currently i am passing input to do fuzzy search.
I have suggested to use Bing Spell Check API, but it is not approved as they think our input may be stored outside.
Is there any option available in azure search to correct the words like "travelexpense".
Is there any option available for this scenario?
The closest I would say is a phonetic Analyzer.
https://learn.microsoft.com/en-us/azure/search/index-add-custom-analyzers
There a couple of other things you can try:
Enable Auto Complete and Suggestions (https://learn.microsoft.com/en-us/azure/search/search-autocomplete-tutorial)
Create synonyms (https://learn.microsoft.com/en-us/azure/search/search-synonyms)
Related
I cannot find any option to achieve a verbatim azure/cognitive/bing Web search.
In my case the difference is trying to sift through tens of millions of irrelevant search results to find the 10 results that actually match my query literally.
Even though I am a paying customer, there is no support available. And the API documentation did not help either.
I would think it should be super easy to provide a verbatim search option. Is there one that I did not see?
I checked further and it seems for the Bing Search APIs - +"phrase" works and returns documents containing this phrase at the top. Just add + in front of what you have been trying. Support link is here: https://azure.microsoft.com/en-us/support/plans/.
My application is in MVC and I have a search text box based on azure index, its working fine however I have a scenario where in if I enter "chres Harris" instead of "chris Harris" it should return result "chris Harris" but it returns different result like "bob Harris" and then "chris harris", I want the results to be nearabout the same even if there is a spell mistake, please help will any index scoring profile, parameter boost or something useful ?
As of now there are two ways you can handle spelling mistakes in Azure Search.
Use fuzzy queries with Lucene query language. You can boost relevance of exact matches over fuzzy matches for example, search=term^2 OR term~2.
If you deal with names of things, like in your example, configure your index to support phonetic search. Different boosting options to influence relevance are described in the article.
Let me know if none of them don't works for you.
The Azure search api offers a fuzzy paramter for suggestions. like this:
https://blssuggestions.search.windows.net/indexes/cities/docs/suggest?api-version=2015-02-28&suggesterName=default&fuzzy=true&search=berlen
Would return "Berlin" as a result of berlen.
I can't find a documentation about this how to activate it in a normal search
setting there fuzzy = true seems to not change anything
https://blssuggestions.search.windows.net/indexes/cities/docs?api-version=2015-02-28&search=berlen&fuzzy=true
[Update]: Please see the other responsed about using querytype=full as this response is no longer correct.
This is correct. Fuzzy search is only available currently in the suggestions api.
You need to call:
https://blssuggestions.search.windows.net/indexes/cities/docs/suggest?api-version=2015-02-28&suggesterName=default&queryType=full&search=berlen~
You were missing querytype=full and the tilde after the character that you want to execute fuzzy searches on.
This is now in the preview version of the api:
https://{yourSite}.search.windows.net/indexes/{yourIndex}/docs?search={fieldToSearch}:{lookupValue}~&queryType=Full&api-version=2015-02-28-preview
Note the ~ and queryType=Full, both of which are required to force fuzzy matching.
Documentation is here:
https://msdn.microsoft.com/library/azure/mt589323.aspx
CAVEAT: The fuzzy search is very fuzzy! i.e. dog will match any 3 letter word with only a single matched letter - dim, now, bag
I am trying to figure out how to tune and tweak but as it is still in preview the documentation is sparse.
UPDATE: I just re-read the documentation and it has since been updated with details of an optional distance parameter. I will investigate.
Suppose I have a huge set of noisy phrases. For each one of them, I want to check if it is defined by some resources by using the google define feature. Once I type "define my_phrase" to the google search box, if the retrieved results contain the definition panel (e.g. https://www.google.com/#q=define+home+cooking), I put it into my phrase pool.
I'm wondering is this possible to do this task in a batch so that I don't have to type each of the phrase manually one by one? It would be great if this could be achieved from a unix terminal but windows is also welcome!
I heard of google-app-engine but I only have a rough idea and not sure if it could help.
Thanks!
as starting point, you may try and play with the Google Custom search following API reference - Xml results
https://developers.google.com/custom-search/docs/xml_results?hl=en&csw=1#XML_Results
Be aware of:
google TOS for this service
quantity courtesy limit
We need to create search input field like it is on _http://maps.google.com
The key functionality is suggest list with appropriate results. We
have not found this feature in API.
Analyzing maps.google.com we see that suggest list is received
from get request to this url
https://maps-api-ssl.google.com/maps/suggest?q=%D0%BC%D0%BE%D1%81&cp=...
There are many parameters, including data from search field. This get
request returns our suggest list.
Is there a possibility to use this url in our needs with our data. Or
how can we make it in some other way.
Similar to our needs: _http://cdn.michaelhart.me/mh/instant/maps/
check this out:
http://tech.cibul.net/geocode-with-google-maps-api-v3/
Theoretically you shouldn't use maps-api-ssl.google.com/maps/suggest as it might not be legal. I found this quote from google employee:
'Endpoints like this that are used by Google Maps but not documented as
part of the Maps API should be considered private interfaces.
Consequently use of those end points is a breach of the Terms of
Service. In addition any existing API credentials you may have are
completely unrelated to these end points because they are not served
by API infrastructure'