To follow up from an earlier blog looking at why search sucks on so many websites, it is probably worth taking a look at what makes a good site search and how this can be achieved.

So what makes a good search site?

  • Language – it is important to use statistical data model that uses mathematical definitions to index data. This means that without having to actually understand the actual meaning of each word, an understanding has been created based on the statistical importance of the word. Furthermore users don’t need to know how to structure their search query as the search engine will work this out based on statistical importance.
  • Spelling mistakes and typos – to solve the problem of misspelling, two algorithms (tri-gram analysis and Levensthein edit-distance) can be used. Tri-gram analysis breaks the misspelled words into blocks of characters and tries to work out how to correct the word in relation to what it knows about the words in the index, using edit-distance as a measure of how misspelled the word is. Together, these two algorithms provide an efficient spell-correction capacity.
  • Naïve search results – using a statistical method can solve a number of relevancy problems. Indexed data uses information about the distribution of terms within the data set to create a weighting system that is applied automatically to rank search results.
  • Field weightings – search result relevance can be greatly improved if field weightings have been applied. Weights are applied to different fields within a product data index.
  • Optimisation and control – being in control is very important for many clients…which can be done by creating a control panel to set the search engine to suit individual needs

Search is becoming more scrutinised by the day and to achieve a great site search it is worth considering the points mentioned and ensuring that the needs of an individual site is taken into consideration.

Please note the views expressed in this blog are the views of the author, Andre Brown and do not represent the view of Locayta, its employees or its shareholders. For more information about Locayta, visit