Ratings and Reputation Spec (Notes)

9/27/2009

Arumble Elements Breakdown

sentiment rating

how strong you feel about the subject

sentiment rating – design

  • structured sentiment is a means of defining a scale for expressing sentiment – this may be grades
  • “adaptability” is important when initially integrating into (or replacing) existing sets of data or methods of conveying structured sentiment already in-place

sentiment rating – context

  • ratings can be related to an article about a topic or combination of topics, where the subject is not concrete.
  • ratings can be relative to a subject that is also an attribute in common among other subjects.
  • ratings can be relative to the “relationship” between two subjects
  • ratings can be relative to an experience related to a subject
  • ratings can be relative to a previous rating

tag

  • words or phrases that describe the subject or attributes of the subject
  • can be ambiguous to attributes, but is unstructured user input
  • use semantics/dictionaries to classify tag meaning as a process for normalization?

tag – relevance

how much/little tag applies to subject

tag – sentiment

how strong you feel about the attribute of the subject the tag describes

tag-attribute inference

  • determine if a tag on a subject actually indicates an attribute of the subject
  • use knowledge of subject design (attribute structure) to link tag to attribute, marked as inference for verification or exclusion in statistical calculation?
  • use semantics/dictionaries to identify relationships

experience indicators

provide way to indicate if tags/sentiment is based on anticipation/expectation or experience with subject

compare subjects

subjects are typically a composite of attributes, some explicitly defined (which can be subjects themselves) and some derived from tags

compare sentiment

compare how the sentiment of one iteration of a subject compares to another, how sentiment changed regarding particular attributes

ranking

rank subjects relative to eachother to indicate “preference”

timeline – trends in tags/sentiment

how do tags or sentiment change over time regarding one subject

timeline – trends in subject iteration

how do tags or sentiment change from one subject/attribute to the next iteration

Arumble Platform Community

engage via point system

  • allow users to gain credibility based on participation
  • allow verification via tools provided to the community that let users flag odd submissions or to thumb up/down content

groupings

lists, such as “best movies of the year” – ranked or unranked (can be ranked by overall ranking/rating or specifically for this grouping)

trend data

n/a

relationships – recommendations

  • allow users to annotate subjects
  • create “Reminds me of…” feature to draw parallels
  • provide tools to compare subjects side-by-side, i.e. such as video clips next to each other to show similarities, or with audio clips or photos or quotes, etc….

9/28/2009

Anticipation rating based on feedback and also anticipation chatter

12/23/2009

Rating styles (structure/orientation)

  • “star” rating scale associated with sentiment words (like Netflix stars)
  • word describing sentiment with “star” rating scale indicating severity of sentiment towards that word (“inverse” of Netflix, stars associated with one word)
  • short 2 or 3-word phrases indicating sentiment toward an aspect or quality of a subject

Rating features (usability)

  • auto-complete for sentiment words, show “tag cloud” of related words and their frequency of use

Leave a Reply