twine 2.x facets
DESCRIPTION
Facets for Twine 2.x, a semantic web application, are a way to change the result set from a search query. This usually just means filtering the results, but with semantic data, this can also mean changing the result set via links of semantic relevance and topics that have affinity to the original query.TRANSCRIPT
Twine 2.x Facets August 12, 2009 James Park
Affinity
FacetsTo better define what facets are, I divided this concept into three other concepts: Filters, Relevance, and Affinity.
RelevanceProvide a series of links related to the search results but not necessarily a match of the search term and filters. These links are easily provided by a semantic index. The links chosen can be subsets or supersets of the search term, optimally just one step away. It is not necessary to visually depict the link’ s hierarchy in relation to the search term for simplicity’ s sake. We used to present these as “Related Searches” links on the Search page.
FiltersProvide a series of actions which takes the current search results and further refines them, removing the items that the user doesn’ t want.
Provide a series of links which have affinity with the search term and filters but is not necessarily related in a semantic context. “Healthy Recipes” could be a context that has affinity with “Core-strengthening Exercises” but not semantically relevant.
FiltersFilters for verticals should be chosen to best match user expectation for the type of results currently being displayed. However, these filters will generally be chosen from a universal set which display its own property in the context of its data type. For instance, alphabetical if it is a string, by time if it is a date, a map if it is a location. There will be cases when the filters should not follow these conventions. One example is a location filter better served by a list of neighborhoods rather than points on a map. Some lists may be better served by dropdowns or other elements.
Chronological
MonthJanuary February March April May June August September October November December
Date RangeFrom
To
Hierarchy
IngredientsMeat Vegetable Spice Asparagus
Bell Pepper Carrot Celery Corn Cucumber
Eggplant Green Bean Leafy Green Onion Pea Tomato
Since we have the semantic information, apply categories to large filter lists. Categories should be kept to a minimum. If there are too many nodes, the filter is probably too broad.
Ranges
Last NameA-D E-G H-J K-M N-P Q-S T-V W-Z
Percentile81%-100% 61%-80% 41%-60% 21%-40% 0%-20%
If the alphabetical or numerical range is known, divide the filter choices accordingly.
Show the best filtering mechanism depending on its time range.
Location
San Francisco, CA
Address
94107
ZIP code
Display the filter in its best visualization representation whenever possible.
Long Tail (too many filters)
ColorBlack Blue Green Indigo Orange Red Violet White Yellow More Auburn
Chartreuse Copper Cyan Gold Gray
Lavender Mauve Periwinkle Silver
NeighborhoodSOMA Civic Center Mission Nob Hill More Bayview
Bernal Heights Castro Chinatown Civic Center Cole Valley
Dogpatch Embarcadero Excelsior Financial Dist. Glen Park Haight
Sometimes there are too many filters and no way to break them up. In this case, display the most prominent ones on the page and the rest in a mouseover pop-up window.
Known Limit
BattingRight Left Switch Other
Batting Average0.301+ 0.201-0.300 0.001-0.200
Number of Children4+ 3 2 1 None
There are cases in which we know the practical limit of a filter. This can mean a finite list or real life constraints. Display only those choices.
Days Hours
Filters 2Here are a few different mechanisms which filters may use, depending on the situation.
Picker
Date RangeFrom
To
Pop-up Window
IngredientsMeat Vegetable Spice Asparagus
Bell Pepper Carrot Celery Corn Cucumber
Eggplant Green Bean Leafy Green Onion Pea Tomato
When hovering the cursor over a choice, a pop-up window appears with more choices appears. We will most likely have to constrain the box size and allow scrolling within.
Links
Last NameA-D E-G H-J K-M N-P Q-S T-V W-Z
Percentile81%-100% 61%-80% 41%-60% 21%-40% 0%-20%
A list of links. Whatever is chosen constrains the search result accordingly.
Pick the filter from a widget formatted for specific selection or over a range.
Select On Map Or Graphic
LocationWash
Last Name
Provide a way to select an area or region on a map or graphic.
Text EntryUser enters in exactly what they wish to filter by. Can be coupled with other mechanisms.
94107
ZIP code
Type-aheadProvide a text box that reduces the number of choices.
Washburn Washer Washington Washman
Days Hours
RelevanceProvide a series of links related to the search results but not necessarily a match of the search term and filters. These links are easily provided by a semantic index. The links chosen can be subsets or supersets of the search term, optimally just one step away. It is not necessary to visually depict the link’ s hierarchy in relation to the search term for simplicity’ s sake. We used to present these as “Related Searches” links on the Search page.
Wheat
Gluten-free recipes
Gluten
Coeliac disease
AllergyStarch
Bread
Related Contexts to “Wheat (food)”Bread food Coeliac Disease disease Gluten protein Gluten-Free Recipes recipe Starch polymer Wheat Allergy disease
AffinityProvide a series of links which have affinity with the search term and filters but is not necessarily related in a semantic context. One way we can generate these links is by scanning the search result set and seeing what tags, categories, and types appear and displaying those. Since these search results are our source for Affinity links, we could keep these with the individual items and make them more prominent as such, or abstract the links into their own listing somewhere else on the Search page.
Wheat Other Contexts Present for “Wheat (food)”Banana food Chicago Board of Trade company Commodity business concept Maize food Multi Commodity Exchange company New Delhi place New York Mercantile Exchange company Sugar food