The Link Optimizer allows fine-tuning the results of the suggested articles and/or products that are generated for each text selection.
If you want to exclude some articles from the Link Optimizer in general, this can also be achieved with a special tag, as described in Data ingestion and indexing.
More fine-grained rules can be achieved via custom profiles. Profiles can be selected and configured using the dropdown on the top right, as shown below:
By default, no profiles are defined, and a default profile is used instead. This might not be optimal for your own linking strategy.
On the profiles page, you can inspect and modify the list of available profiles.
A profile is mainly composed of two parts:
- Article level parameters: used for automatic link generation on the whole text article. These are configured via the Link Selection Configuration.
- Selection level parameters: used for automatic and manual link generation on an individual text selection. They are configured via the Article/Product Recommendation Configuration.
This section gives an overview of the different parameters found in each profile. This information is also available directly on the page in tooltips that appear when hovering over the names of the parameters.
Link Selection Configuration (used for both Article and Product Linking)
This only applies to automated link generation using the "Generate Links" button. Manual links are not affected.
- Average Links per Paragraph is used to compute the total number of links that will be generated for the article based on the number of available paragraphs. (Note: existing external/manual links will be subtracted from this total, so that it may happen that no new links are generated)
- Articles/Products per Selection is used to set the number of suggested articles/products to choose from for a given selection (higher = more choices, but it also takes longer)
- Preferred Selection Length / Selection Length Weight: are used to influence the number of words composing the candidate selections.
- Selection to Source Article Weight is used to favor selection(s) being similar to the content of the source article
- (Product Linking only) Include Generic Products decides whether generic products, like just "iPhone", should be included in automatic product link generation. Disabled by default.
Article Recommendation Configuration (used for Article Linking)
- Date Range: is used to filter the suggested articles to no older than a given amount of days (measured by their last-updated-timestamp). When set to zero ("No Filter"), this filter is disabled, and articles of all ages can be suggested.
- Recency Weight: is used (independently of the date range hard filter) to slightly alter the recommendation results to include more recently modified articles when set to a high number. When set to zero, articles of all ages are recommended evenly.
- Category / Relevance: are used to filter the recommended articles by their category. When the relevance is set to maximum ("Required"), this is a hard filter. When set to a lower value, articles of the selected categories are favored over others as recommendation results, but articles of other categories might still be suggested when they are deemed a good fit due to their other properties.
- Premium: can be used to filter the suggested articles based on their public accessibility: Profiles can be set to only return publicly accessible articles, or only premium articles, or both.
- Similarity Weights: (Similarity to Selection / Paragraph / Article) are used to take into account the similarity of the source selection / paragraph / article to the target article. By adjusting these three weights, we are able to take into account the context of the selection at various degrees. For example, if the text of a given selection is part of a very popular content, it might be useful to increase either the paragraph or article weight so that the suggested results will be closer to the context described in the source paragraph or article.
- Similarity Confidence: defines how important the match quality between anchor text and suggested recommendations should be. Lower values are better for exploration purposes, while higher values are better to have a high certainty of good matches.
Product Recommendation Configuration (used for Product Linking)
- Minimum Price / Maximum Price: are used as hard filters to limit the recommended products to be in a certain price range.
- Minimum User Rating: hard filter for how many stars (out of 5) users shall have given the recommended products on average. The maximum is "4", as no product can have a perfect 5/5.
- Prime Only: Only Amazon products with the Prime Badge will be recommended. Disabled by default.
- Associate Tag: can be used to define a custom Amazon associate tag to be used for creating product links, instead of the configured default tag.
- Set profile as default: can be checked to use this profile per default in the future. The default article linking / product linking profiles are pre-selected when an article is opened the first time in the Link Optimizer, or when the profile that was previously used for an article (or a single link) was deleted.
Custom Profiles for individual selection
During automated link generation, only a single profile can be used as main strategy to rank which articles and/or products will be finally suggested. However, in certain cases, it may be required to apply multiple strategies to different subsets of selections.
For these purposes, each individual selection can be customized by setting a separate profile and regenerating the recommendations just for this selection.
Inside a link selection, the associated profile can be changed by clicking on the settings button, as shown below:
You can then change the profile selected to any other available profile, as shown below: