Recommender
Data Harmony® has built the Recommender to suggest additional items which may be of interest to the searcher based on the items already retrieved. Once the user has selected items of interest the Recommender comes into play. There are many theories about how to create recommendation engine, and most depend on the user profile. We don't do it that way. In our case we already tag the information objects with terms from the taxonomy. Therefore, the easiest way to make that selection is to choose another item with exactly the same index terms as the one the user selected. If the data has been tagged with taxonomy terms then the more closely the indexing matches, the more likely the recommended articles will satisfy the user. That is, if articles get 5 index terms each from the taxonomy attached to them and if the search likes one of the articles then another article which has the same five index terms will most likely also answer their needs. We believe that exactly the same indexing would give a very good recommendation to the user. When there is no 100% match we go to the next highest level of matches. Confidence as given by relevance ranking is then an absolute confidence factor and not a computer system guess based on recent search profiles.
The Data Harmony Recommender is dependant on having access to the taxonomy terms applied to the individual documents. As with the rest of Data Harmony, the Recommender is repository and search software agnostic. Because it is not using co-occurrence but rather the keywords associated with the individual document, Recommender is a less system intensive and more precise way to produce relevant results to the searcher.
To see the Recommender in action try the link below First enter a search term into the box, then click on one of the hits; it will show you other items with the same indexing. For e-commence implementations we add the suggestions to the selection page.
