Ravelry.com: Examining Classification and Categorization

1.     Does your displayed index have subheadings? About how many levels of subheadings (depth) does a typical heading have? What's the deepest number of levels of subheadings you can find? Give examples.

Yes; the displayed index has subheadings. The typical subheading depth is 1 level, but 2 level depth is common as well; the deepest are 3 levels.

Example 1 level subheadings:

      Heading: Fiber

                  Subheading: Acrylic

                  Subheading: Alpaca

                  Subheading: Angora

                  Subheading Bamboo…

 

      Heading: Weight

                  Subheading: Thread

                  Subheading: Cobweb

                  Subheading: Lace…

 

Example 2 level subheadings:

      Heading: Attributes

                  Subheading: Construction

                                          Subheading II: Intarsia

                                          Subheading II: corrugated ribbing

                                          Subheading II: mosaic…

Heading: Needle Size

                  Subheading: Small Needles

                                          Subheading II: US 000 – 1.5 mm

                                          Subheading II: US 4/0 – 1.25 mm

                                          Subheading II: US 5/0 – 1.0 mm

 

Example 3 level subheading:

      Heading: Category

                  Subheading: Clothing

                                          Subheading II: Sweater

                                                                  Subheading III: Cardigan

                                                                  Subheading III: Pullover…

                                          Subheading II: Tops

                                                                  Subheading III: Sleeveless

                                                                  Subheading III: Strapless

                                                                  Subheading III: Tee…

2.     Does your index has a restricted set of top level categories intended to convey the subject scope (i.e., the top level categories encompass all subcategories in the classification scheme and all documents that might be added to the collection)? Explain by giving examples of the top level categories and some subheadings.

Yes, there are a restricted set of top level categories that shape the scope of the collection.

Example 1: This top level category encompasses all subcategories of pattern classification, and limits the collection to include documentary units with pattern information.

Top Level Category: Patterns

            Heading: Craft

                                    Subheading: Crochet

                                    Subheading: Knitting

                                    Subheading: Machine Knitting

                                    Subheading: Loom Knitting

Example 2: This top level category encompasses all subcategories of project classification, and limits the collection to include documentary units with project information.

Top Level Category: Projects

            Heading: Pattern Category

                        Subheading: Clothing

                        Subheading: Accessories

                        Subheading: Home

                        Subheading: Toy / Hobbies

                        Subheading: Pet

                        Subheading: Components

3.     Are there consistent semantically meaningful relationships or associative relationships between headings and immediate subheadings in your index? If so, what types of relationships hold between headings and subheadings? Are the relationships between headings and subheading clear and consistent or is it hard for the user to intuit these relations. Give positive or negative examples.

Overall, there are consistent semantically meaningful relationships. This is likely the result of the programming language used to design the taxonomy, which requires consistent semantic relationships to allow for meaningful computation. However, there are instances where some associative relationships have developed.

Example 1: Semantically consistent:

            Heading: Craft

                                    Subheading: Crochet

                                    Subheading: Knitting

                                    Subheading: Machine Knitting

                                    Subheading: Loom Knitting

Each of the subheadings have an is:a relationship: Crochet is a craft, Knitting is a craft, and so on.

The is:a relationship is the one which founds all relationships on Ravelry.com, and it is a consistent and common association which allows users to easily intuit the relationship across headings.

However, there are still some associative relationships within this structure. Consider:

Example 2: Associative relations:

      Heading: Category

                  Subheading: Clothing

                                          Subheading II: Shrug / Bolero

                                          Subheading II: Sweater

                                                                  Subheading III: Cardigan

                                                                  Subheading III: Other

                                          Subheading II: Tops

                                          Subheading II: Other

The bolded elements are those which the subheading breaks the is:a relationship; the italicized element is one that breaks a subheading relationship depending on your understanding of the term.

“Other” is not a sweater, though a sweater is a type of clothing. Similarly, “Other” is not a top, though a top is a type of clothing. This subheading is not particularly damaging to user navigation because the overwhelming majority of subheadings are clear and consistent and the documentary units in each category are small, however; it is worth noting that any “misc.” category such as “Other” usually indicates a need for additional, clearer subheadings.

In the case of “Shrug/Bolero,” it is somewhat disruptive to the parent/child relationship for the “Shrug/Bolero” to be a parent with no children, while “Sweater” is the parent to comparable items such as “Cardigan.” I would argue that colloquially speaking, most English users would associate a shrug to sweaters as they do a cardigan. This too, however, is relatively undisruptive because the is:a relationship is not broken between shrug/bolero and the subheading “clothing.” A shrug is a type of clothing, so users can still easily navigate to it with relative ease.

4.     Are 'sister' categories (ones that share the same parent/heading) complete, distinctive and mutually exclusive? Give positive or negative examples.

As discussed in question 3, there are some sister categories that share a parent and include “other” as a subheading:

Heading: Category

                  Subheading: Clothing

Subheading II: Sweater

                                                                  Subheading III: Cardigan

                                                                  Subheading III: Other

 

With regard to completeness, categories which include “other” are likely incomplete, as they do not address a specific term with which to identify the documentary units inside.

There are, however, numerous sister categories that are complete/distinctive/mutually exclusive:

            Heading: Needle Size

                        Subheading: US 00 – 1.75 mm

                        Subheading: US 0 – 2.0 mm…

Needle Size is an excellent example, as each subheading is a distinct, mutually exclusive size (a US 00 needle cannot ever be a US 0 needle), and the collection of subheadings contains all manufactured sizes of needles available to users. “Completeness” could be argued should any new sizes be introduced, but for the purposes of the collection, the size range is complete.

5.     To what extent is alphabetization used to help users browse displayed indexes?

With rare exception, all subheadings to a heading are alphabetized. In the case of “Has Photo,” the subheading “Yes” appears before “No,” but the size of this tree (only 2 subheadings) is not any less navigable as a result of this.

6.     If your index uses faceted classification, what facets does it use and how are they ordered?  Are the facets distinct and complete? Provide examples of facets.

Although Ravelry.com is not entirely faceted classification, it does use facets for its heading structures within top level categories. In the case of Patterns (a top level category); the facets are:

Craft, Category, Availability, Has Photo, Attributes, Gender/Age/Size/Fit, Weight, Yardage, Meterage, Colors Used, Source Type, Hook size, Fiber, Needle Size, Rating, and Language.

As facets of Patterns, they are distinct and complete. All patterns will address these facets; a pattern that does not include hook or needle size is effectively not a pattern. Some facets are more important than others to locate a documentary unit of interest, and not all facets must be addressed to access a documentary unit.

Consider, the Craft and Category facets are ordered first. The Craft facet addresses what kind of pattern the user would like to use (knit, or crochet) and the Category facet addresses what kind of product the user would like to produce (clothing, toys, etc.). These are the predominant determining factors to a users’ pattern selection, so it is reasonable to order them first.

7.     If your index has neither top level categories nor faceted classification, explain how the index terms are displayed/arranged. For example, an index might have a combination of a traditional classification schema plus a set of filters (facets) for characteristics of all or specific subsets of resources. This is particularly common in taxonomies.

Although I have presented examples of top level categories and facets, Ravelry.com ultimately displays and arranges index terms using a combination of traditional classification strategies and facetted classification; it is a great example of a blended taxonomic classification.

As discussed in question 6, the facets of each top level category are displayed in order of relevance to the user. In the case of Ravelry, “relevance” can be defined as the facets that most quickly filter a top level category. By organizing facets by their “strength” to filter results, the scheme allows users to browse at high speeds. The intended users of Ravelry.com are customers, so the classification scheme is optimized to allow the user to refine the documentary units displayed to them very quickly: they can arrive at a unit of interest and make a purchase (or encounter an ad along the way and make a purchase at an affiliated site).

8.     What impact does the classification system (or lack thereof) have on the usability and maintainability of the index? For example, assess your index's classification schema with respect to predictability and collocation. If there are no top level categories or few subheadings, what is the effect of the absence top level categories or of subheadings on the index usability and maintainability? What kinds of top level categories or subheadings or facets could be usefully added? Feel free to discuss any other aspects of the classification of your index relevant to this assignment.

Similar to the usability and maintainability concerns I addressed in Scope, the entities involved in indexing Ravelry.com are usually established products: clothing, accessories, or soft furnishings. There may be ambiguity in some entries, but most documentary units will be straightforward to identify.  As a result, the blended structure of top level categories and facets with headings/subheadings is efficient at allowing users to use the collection. They are able to filter quickly, and arrive at documentary units of interests with relative ease because their primary concerns (the facets) are addressed in order of importance.

Additionally, the classification system has established facets to use. If the facets were more ambiguous, the index would be less usable and maintainable – it might very quickly contain facets beyond what would support a successful browsing experience. However, knitting and crochet have very established facets (such as products created or techniques used) and this limits the facets to a navigable degree.

Where there are “Other” subheadings, I think there is some room for additional research to increase usability. If I want to access a “tube-top pattern” (of which there are numerous documentary units), it would be reasonable to establish this as a subheading to “tops” rather than lumping into “other.” Overall, however, I believe Ravelry.com has a very consistent and navigable classification structure for its intended users, and will be able to maintain the index well with the structure they are using.