For up-to-date information, please follow this link.
Introduction
Cxense customers can apply a standard Cxense taxonomy and/or any number of custom taxonomies to the processing of their web page content. The purpose of applying any taxonomy to the content processing is to improve the identification of each user's interests. If applied correctly, users will end up being tagged with keywords that match their short and long term interests.
A taxonomy is a set of key-value pair mappings. The keys are terms that can be found in the documents (web pages) such as for instance Toyota and spaghetti bolognese, and the values are taxonomy paths that we use for tagging the document, such as motor/cars/toyota and food/pasta respectively to complete the two examples we started with. Users that then read documents tagged with for instance food/pasta will then have that tag added to the information Cxense stores on them. A complete list of all the paths in the standard Cxense taxonomy is found here.
The taxonomy key-value pairs are maintained in a dictionary that can be in the form of an Excel file or JSON structure. The practicalities of administrating and updating any kind of custom dictionary (taxonomy or linguistic) is explained in Dictionary Management Tutorial. Customers can either rely on the standard Cxense taxonomy in which case they don't have to do anything, or they can in addition or instead make use of their own custom taxonomy which they then have to maintain.
As of August 2016, the Cxense standard taxonomy comes in the following languages: Danish, English, Finnish, Indonesian, Norwegian, Russian, Spanish and Swedish. If the standard Cxense taxonomy is enabled and the detected language of a document is not among the supported ones, no taxonomy tagging will take place.
WARNING: Don't Confuse URL Based Content Tagging with Taxonomy
There are two content profile fields that can easily be confused for having something to do with taxonomies without having it:
-
taxonomy - The taxonomy field that Cxense Insight creates and maintain for each web page tracked, does not have anything to do with a traditional taxonomy per se. The value of the taxonomy field is the path of the page url. Hence, a web page with the url http://server.domain/sports/football/some-football-page.html would hold the value sports/football. This may look very much like a taxonomy path and in many cases it will do an excellent job as a taxonomy path, but in reality it is not. It is simply a part of the url path.
-
category - The category field will by default contain a list of all the path fragments from the so-called taxonomy field. Hence, in the particular case of the example above, the category field will hold a list with the two values sports and football.
The field that will hold the actual taxonomy paths is xyz-categories (where xyz is a placeholder for the customer prefix).
How does the Taxonomy Feature work?
The figure below shows how the taxonomy feature works whether it is a standard Cxense or a customer taxonomy. The blue steps represent customer actions, the green ones are done by the system. The first thing that needs to be done for any taxonomy to be used are blue steps 1 and/or 2. Blue step 1 enables the use of the standard taxonomies for document written in any of the supported languages. Blue step 2 would be as shown in the figure, the uploading of an Excel with a customer made taxonomy (the use of JSON files is also supported).
Unknown Attachment
Blue step 3 and green step 3 represent the two alternative ways of uploading the data that is to be processed. In most cases this will be done by the Cxense crawler (green alternative). However it is also possible for the customer to upload data using the API (blue alternative). In either case, the taxonomies will effect the uploaded data as shown in common green step 4 where the uploaded taxonomy key-value pairs come to play. In the figure we have imagined that the two terms Toyota and spaghetti bolognese appear in the document in a such a way that they end up being picked. The xyz-categories attribute of the content config will then have the taxonomy paths motor/cars/toyota and food/pasta added to the list of paths already held by the attribute (xyz is here just a placeholder for the actual customer prefix).
Enabling the Standard Cxense Taxonomy
Speak with your Cxense onboarder or send an e-mail request to support@piano.io if you wish to enable the standard taxonomy for your site(s). The document Enabling the Standard Piano Audience Taxonomy describes how to do the task (link only accessible to Cxense employees).
Uploading a Custom Taxonomy
Only a Cxense representative can enable the use of a custom taxonomy, but customers themselves are expected to maintain them. The Dictionary Management Tutorial explains both how someone from Cxense can enable the dictionary and how a customer can update it.
Below we have made a taxonomy dictionary template out of the standard Cxense taxonomy. Customers that are to make their own custom dictionaries can use it (or the actual taxonomy Excel file) as an starting point.
{
"global-properties": {
"tokenizer-context": "<THE TARGET LANGUAGE: en, fr, es, de, no, sv, etc.>",
"mode": "overlap",
"leftmost-longest-match": true,
"key-normalization-flags": 4,
"value-normalization-flags": 4,
"swap-fields": true,
"count": 2,
"unique-count": 2,
"count-field": "value",
"expand-paths": true,
"group-prefix": "<THE CUSTOMER PREFIX>"
"annotate-paths": false,
},
"food": {
"spaghetti bolognese": "food/pasta",
"bibimbap": "food/asian/korean"
},
"motor": {
"toyota": "motor/cars/toyota",
"renault": "motor/cars/renault",
}
}
Combining Standard and Custom Taxonomy
It is possible to combine the use of the standard Cxense taxonomy with one or more custom taxonomies. The only caveat to take into consideration is when the same key-value pair appears in more than one of the taxonomies. The weight of the taxonomy value will be the sum of the contributions from each taxonomy. This is something that one more often than not probably would like to avoid by making sure one doesn't have duplicate key-value pairs across the taxonomies.