We present automated algorithms that enact three strategies used by educators and journalists to make unfamiliar measurements more understandable:

• Adding Familiar Context

  Presents a measurement (e.g., 28 lbs) alongside objects with similar measurements (e.g., the weight of a tool box, the weight of a microwave).

• Reunitization

  Re-expresses a measurement (e.g., 3 ft) using a more familiar object as the unit with a multiplicative scale factor for converting from one to the other (e.g., 2 times the height of a single bed).

• Proportional Analogy:

  Re-expresses a pair of measurements (e.g., the ratio between the volumes of Mercury and Earth) using two familiar objects that have measurements with the same ratio (e.g., the ratio between the volume of a sugar bowl and a watering can).

We devise a scalable approach to mining object datasets comprised of three stages:

• Identify familiar classes of objects

  with WordNet, ImageNet, and crowdsourcing. We first identify criteria for effective re-expression objects like Concreteness, Countability, Rigidness, Low Measure Variance, and Familiarity.

• Collect instances of the objects

  from online resources like the Amazon product database and Wikipedia

• Filter false positives and collect familiarity data

  using crowdsourcing

Abstract: It can be difficult to understand physical measurements (e.g., 28 lb, 600 gallons) that appear in news stories, data reports, and other documents. We develop tools that automatically re-express unfamiliar measurements using the measurements of familiar objects. Our work makes three contributions: (1) we identify effectiveness criteria for objects used in concrete measurement re-expressions; (2) we operationalize these criteria in a scalable method for mining a large dataset of concrete familiar objects with their physical dimensions from Amazon and Wikipedia; and (3) we develop automated concrete reexpression tools that implement three common re-expression strategies (adding familiar context, reunitization and proportional analogy) as energy minimization algorithms. Crowdsourced evaluations of our tools indicate that people find news articles with re-expressions more helpful and re-expressions help them to better estimate new measurements.

We describe a scalable approach for mining object databases (Amazon, DBpedia) to construct a database of familiar objects and their measurements, using semantic databases (WordNet and ImageNet) and crowdsourcing techniques. We present automated re-expression algorithms that apply the data in this database to re-express unfamiliar measurements in terms of more familiar objects using Reunitization, Adding Familiar Context, and Proportional Analogy strategies. We find that our tools improve measurement estimation and the perceived value of article content involving measurements.