1. This paper presents automated methods for extracting logically related motions from a
 data set and converting them into an intuitively parameterized space of motions.

2. Large motion data sets often contain many variants of the same kind of motion, but 
without appropriate tools it is difficult to fully exploit this fact.

3. Challengers
(1) Logically similar motions can have very different skeletal poses, but existing 
computational methods can only reliably find motions that are numerically similar to the 
query in the sense that corresponding skeletal poses are roughly the same.

4. Key points
(1) To find logically similar motions that are numerically dissimilar, this paper presents
 a novel search method that uses numerically similar matches as intermediaries to find 
more distant matches, together with a precomputed representation of all possibly similar 
motion segments that makes this approach efficient.
a. Multi-step search.
b. Using time correspondences to determine similarity.
c. Interactivity through precomputed match web.

(2) This paper presents an automatic procedure for parameterizing a space of blends 
according to user-specified motion features. This algorithm samples blends to build an 
accurate approximation of the map from motion parameters to blend weights, and it uses a 
scalable scattered data interpolation method that preserves constraints on blend weights.

5. Limitations
(1) Not all logically similar motion can be found. 
(2) Motion sets found by the search engine are not guaranteed to be blendable.