Spelling support in SiteSeeker

11/1/2012

Kunskapsdatabasen, Introduktion

SiteSeeker can give the visitors spelling suggestions. The spelling suggestions are presented when SiteSeeker has found another spelling that producers more or better results, i.e. not only when there are no hits at all for a certain query.

How does it work?

The spelling support is not a spell checking program; it only suggests words that really exists in any of the searchable documents. There is a big advantage in suggesting occurring words only, since many unusual proper names does not exist in dictionaries. A disadvantage by using the same method is that SiteSeeker can suggest a misspelled word if such a word occurs in an indexed document.Så här kan det se ut om man stavar fel på denna webbplats

When SiteSeeker presents related words, a "distance" is calculated between the word found in the search query and all words that exist in the indexed documents. The distance is shorter if letters that separates the two words sounds similar (c and k, dokuments), are close to each other on the keyboard (v, b, lettwrs) and if they can be mistaken from a appearance perspective (b and d, bocuments). In addition, the distance shrinks if letters are missing, have been added or have changed place (docments, documments, documnets). Double consonants and if the difference is located in the end of the word will also create a shorter distance (documentds).

By taking into consideration all these types of typographical errors, the spelling support can assist in correcting misspellings and also assist in spelling proper names. It is also a great help for dyslectics and people with English (or another language used on a website) as their second language. SiteSeeker will evaluate all suggestions internally and will present only those that will produce the best result, qualitatively as well as quantitatively.

The spelling support is set to allow alternative search terms that are seemingly far from the word you are typing. The configuration will always try to balance between how often "correct" results are displayed and how often "incorrect" results are shown. A configuration with less incorrect results will also mean greater risk to miss out on the correct results.

How do you enable the spelling support?

The spelling support is integrated in SiteSeeker and does not require any special actions in order to enable it.Please note that if you are using an integration module, for example SiteSeeker for EPiServer, you need to make sure the spelling suggestions are not concealed in the interface. Test the functionality by typing a misspelled search query here on our website to see firsthand how we have chosen to design the spelling suggestions!

There are some settings you can make that affects the spelling support. For example, in Admin, you can set how "close" the suggested word should be (see discussion regarding distance between related words earlier in this text). There is also the option to turn off the spelling support altogether, even though we would advise against it. You find the settings in Search pages.

If the spelling support suggests a word which you think should not be suggested, use these alternatives to delete the suggestion:

  • If the suggested word is a misspelling. When the misspelled word has been corrected in the source document, the word will no longer be suggested after the next indexing.
  • If the search word is correctly spelled, but does not produce results if the search word is relevant for the documents on the website it can either be written in the document, alternatively in its meta information (keywords or description) if it should not be in the text. Yet another method is to add a synonym or to create a best bet.
  • If the suggested word is inappropriate By writing inappropriate words in a blocked style (i n a p p r o p r i a t e) or by inserting <!--eri-no-index--> and <!--/eri-no-index--> around the word, you make sure that SiteSeeker does not index the word and it will then not be suggested (or produce a result when querying).