Continuing the journey into Phrase Based Optimization
One thing worth mentioning, is that there is limited info relating to personalized search and PaIR. It merely touches the surface of the over-all personalized search methodologies. This means it would merely play a role in the PS engine. There is much more to it and the PaIR model aspects are by no means comprehensive. I simply wanted to give a quick break down as to how a PaIR system would handle PS processes.
Personalized Search in a PaIR system looks to customize the ranking of the search results based on a perceived model of a user's particular interests. Information deemed to be relevant to the user's interests would rank higher in the search results.
The ‘user model’ is defined via search queries and pages, both of which can be represented by phrases. The user query – along with associated extensions – makes up the first part of the profile. Next, the page is analyzed for common (or ‘good) phrases associated with the page (the Body Score). At this point, the ‘user profile’ can be represented as the culmination of a set of queries with a set of documents/web pages. The pages included in the profile can be weighted from results the user selects in SERPs, using “a client-side tool which monitors user actions and destinations”; which in all likelihood would be a cookie from the ‘Google Account’ system. To quote – “where a user is recognized by a login or by browser cookies”
|“ A client side browser tool monitors which of the documents in the search results the user accesses, e.g., by clicking on the document link in the search results. These accessed documents for the basis for selecting which phrases will become part of the user model. For each such accessed document, the search system 120 retrieves the document model for the document, which is a list of phrases related to the document. Each phrase that is related to the accessed document is added to the user model.”|
By looking at what documents the user selects out of the predictive PaIR model of query results, it can further refine the relevance of the model in a unique ‘user model’. Each document that is selected from the given query results (SERP) has its ‘phrase profile’ added into the weighting of existing phrases in the user model. Each one a further refinement from the last.
Furthermore, it could be tuned to add weighting to user actions on a given document such as;
- stored as a favorite
- stored as a link
- email to another user
- maintained open in a browser window for an extended period
These could be perceived as a ‘higher level of interest’ and as such given greater weight in the stored phrasing profile of a given user.
The personalization aspect is applied PRIOR to the a standard PaIR implementation, Meaning; if the results from the query term DO NOT pass the inspection of the ‘user model’ then it is unlikely that it will rank high for the given query - Though not outside of the realm of possibility.
The system also comes into play with ‘Personalized Topic Based Document Descriptions’ – which is essentially seeking to also tailor the query results descriptions ( or snippets) to the user model. The method compares the query related phrases with the user model of phrases in high frequency in accessed documents. It would use the resulting set of results as the basis for ranking sentences in pages of the results.
It ranks the sentences based on the same relational factors of the query and user model phrases. The resulting description is comprised of the highest ranking sentences of the query result documents. It hopes to represent ‘the key sentences of the document that express the concepts and topics most relevant to the user’.
If you have been following along the PaIR road at home, much of this should be getting easier by this point. The Personalized Search methods further highlight a site ‘theme’ approach. It is s secondary layer added to the process. It’s in essence, a second sweep of the system prior to ranking documents in the traditional methodology.
Until next time – keep it ‘relative’