Contents

(+ show- hide)

This page provides information about the review opinions data set. An opinion is a fragment of a review (or sentence) that contains a mention of a feature (for example, size) with a given sentiment (for example, positive sentiment).

Bazaarvoice's sentiment analysis uses two natural language processing techniques: double propagation and aspect-based sentiment analysis. These techniques allow us to break down individual sentences into multiple opinions ("I love the picture quality || but hate the sound quality") and analyze the sentiment of those opinions independently. This segmentation is crucial because consumers frequently express multiple opinions in the same sentence. Without a way to capture these opinions, a significant amount of customer experience data is lost or interpreted inaccurately.

An opinion has five components as detailed in the example response below:

  • Entity - a product or a service (for example, a smart phone).
  • Feature - a feature of the entity (for example, a touchscreen on a smart phone).
  • Sentiment - a measure of the sentiment towards the feature: positive, neutral, or negative. These values are represented as 1, 0, and -1 in the data tables.
  • Holder - the review author who created the opinion.
  • Time - the review date when the opinion was expressed.
For products with more than 1000 reviews, only reviews from the last two years will be incorporated into opinions.
Products must have a minimum of 10 reviews to have product sentiment insights created.

Example

The following is a generalized example. Continue to the Fields section for information on each field, or view the Workflow Walk-Through tutorial to learn how to request this data set.

{
  "id" : "oxclOi9PeKpJFKgmu6aySQQ0V//9HIX8",
  "feature": "appearance", 
  "sentiment": 1,
  "text": "This Flatware set is just so stylish and attractive",
  "authorId":  "zt9g5w56a39kwiog0nui8wu1zk",
  "reviewId": "026f8cfe-30d3-5f6f-8f63-69078572833c",
  "reviewLegacyId": "111119566",
  "reviewDate": "2017-12-16", 
  "productName": "45 pc. Flatware Set with Buffet Server",
  "productExternalId": "2891130", 
  "category": "Flatware",
  "client": "acme",
  "categoryId": 1001
}

Fields

id

"id" : "oxclOi9PeKpJFKgmu6aySQQ0V//9HIX8"

The unique ID of the review opinion.


feature

"feature": "appearance"

The aspect of a product that is mentioned within the opinion text. Bazaarvoice uses natural language processing to automatically identify features from review text. There is no fixed feature dictionary and features may differ between categories (for example, “picture quality” is a feature of televisions but not of socks).


sentiment

"sentiment": 1

An integer representing whether the sentiment expressed about the feature is positive (1), neutral (0), or negative (-1).


text

"text": "This Flatware set is just so stylish and attractive"

The analyzed text fragment from the review.


authorId

"authorId":  "zt9g5w56a39kwiog0nui8wu1zk"

The unique identifier of the author.


reviewId

"reviewId": "026f8cfe-30d3-5f6f-8f63-69078572833c"

The unique identifier of the review the text fragment is a part of.


reviewLegacyId

"reviewLegacyId": "111119566",

The unique identifier previously used by Bazaarvoice to identify the review.


reviewDate

"reviewDate": "2017-12-16"

The submission date of the review.


productName

"productName": "45 pc. Flatware Set with Buffet Server"

The name of the product that the review applies to.


productExternalId

"productExternalId": "2891130"

The client's unique identifier for the product.


category

"category": "Flatware"

The category that the reviewed product belongs to.


client

"client": "acme"

The client from which the review is sourced.


categoryId

"categoryId": 1001

The unique identifier for the category.