Faceted Search (Faceted Navigation or Faceted Browsing)

Faceted search (also faceted navigation or faceted browsing) — a method of accessing information using several filters — facets — at the same time. It is organized according to a faceted classification

To use faceted search, each information unit is classified according to several explicit characteristics (facets) . Faceted search is contrasted with the taxonomic order of filtering and ordering .

Facet properties can be obtained by analyzing the text of an information item and extracting characteristics. Or by a set of fields that already existed in the database (for example, author, language, format, reading time). In this way, existing web pages can gain additional navigation through faceted search.

In 2006, a workshop on faceted search was held in Seattle.

Comparison of full-text and faceted search

Full-text search — searching for products by word or phrase. For the user — this is a text input field with a «Search» button, available on any page of the site.

Faceted search — searching for a product by several characteristics: color, size, memory capacity, price, etc. For the user — this is a set of filters. Each filter is linked to only one characteristic and vice versa. The filter's values — all possible values of the characteristic. The user sees filters on the section page, the category page, and on the full-text search results page. When the user selects a value, the filter is considered active.

The difference is that facets let you select only intersecting values, while ordinary filters simply select information according to the entered data.

Faceted search by itself is a narrowing selection, in which the possible filter values depend on the results already present in the selection. That is, if there are filters by brand and OS, then clicking Apple makes the "Android" filter disappear, preventing the user from clicking criteria that would give a zero-result selection.

What, then, are the essential differences between faceted search and simple search?

• First, for filters with numeric ranges, the smallest and largest values are automatically substituted (for example, in the «Price» search filter the «From» and «To» fields will be filled with the smallest and largest price of a product in the category).
• Second, only the available variations will be highlighted in the list of filters (for example, when selecting T-shirts with size L in the filter, the colors for which there are no models of the chosen size will disappear from the color filter).
• Third, when working with filters, the shopper will immediately see how many products match the parameters they set (for example, when selecting size L and white color, the search will display the message "N products found").

You can see an example of faceted search in the screenshot below.

Developing the concept of faceted search

Navigational search	Faceted search

The Association for Computing Machinery, within the framework of SIGIR, presented this description of the role of faceted search:

• Navigational search, using a hierarchical structure (taxonomy), where users sequentially browse the information space, narrowing the search scope in a given order, as in the DMOZ or Yahoo! directory .
• Direct search, using text queries and showing the most relevant (matching) results above the rest. This is how all search engines work: Google, Yandex, Rambler .

SIGIR states that direct search has taken the dominant position, while navigational search is becoming less and less popular . Faceted search lets users move through a multidimensional information space by combining text search with a gradual narrowing of choices along each dimension . It is the dominant mechanism on e-commerce sites . As of 2006, it was extending to semi-structured data and tags (arbitrary labels) .

Behavior of facet filters

Briefly, it works like this: a filter filters products and filters the choices available in other filters.

Filtering products

This part is simple. The user has selected:

1. one value, and sees products matching that value;
2. several values in one filter, and sees products matching at least one of them;
3. values in several filters, and sees products matching a value from each filter.

In terms of Boolean algebra: a logical «AND» applies between filters, and a logical «OR» between values within a filter. Simple logic.

Filtering the choices available in other filters

«Well… whatever options exist — are shown, whatever doesn't — is hidden» — that's roughly how business describes filter behavior. Sounds logical. In practice, it works like this:

1. We go into the Phones section and see filters by characteristics: Brand, Screen size, Memory. Each filter contains values.
2. We select a brand. Some values disappear from the Screen size and Memory filters. In the Brand filter, all values remain as in step 1.
3. We select a screen size. More values disappear from the Memory filter, and some values disappear from the Brand filter. The values in the Screen size filter remain as in step 2.
4. We select memory. More values disappear from the Brand and Screen size filters. The values for the Memory filter remain as in step 3.
5. We «reset» the selected values in the Memory filter. The filters restore the state from step 3, and so on.

The number of filter values depends on the number of products: the more products with a differing value of a characteristic, the more values in the filter. The user reduced the number of products in the selection for the other filters when they chose a brand. This led to the value lists being updated.

From this follows a universal rule: filter values are extracted from the product selection formed by the remaining active filters.

Each active filter has its own product selection.

If we have N filters and:

• none are active, the selection is the general one. It is the same for all filters, and matches the search results;
• M are active, and M < N, then the number of selections is M + 1, where 1 — is the selection to which all active filters are applied. It is the same for all inactive filters and matches the search results;
• M are active, and N = M, then the number of selections is N. Each filter has its own selection.

As a result, when the user selects a value in a facet filter, the following happens:

1. a search selection of products is formed;
2. values for the inactive filters are extracted from the search selection;
3. for each active filter, a new selection is formed, and new values for the active filters are extracted from it.

Use of faceted search in commerce

Faceted search is widely used in commerce, for example in online stores and libraries. More and more developers are providing software solutions for faceted search.

Online retailers typically offer the ability to filter by product type, brand, and price.

In the early 2000s, retailers began using faceted search. A 2014 benchmark of the 50 largest American online stores shows that, despite the advantages of faceted search, only 40% of sites had implemented it

Such a search helps potential buyers avoid getting lost among the whole variety of offers and find exactly what they need.

Librarianship

In 1933, librarian Shiyali Ramamrita Ranganathan proposed a facet classification for books, now known as the Colon classification. However, it failed to replace the more popular Dewey Decimal Classification .

Modern electronic library catalogs are increasingly using faceted search. For example, the Online Computer Library Center, WorldCat, and CiteSeerX .