Name
Abstract | ||
|---|---|---|
Catalyst annotations are standoff (versus inline) which means that the underlying signal is unmodified and anno- tations are maintained and communicated separate from the signal. By separating the signal from the annotations and an- notations of different types from each other, Catalyst is able to automatically construct customized streams of annotation for each component in a system. The set of annotations, attributes and their names can all be transparently modified between each language processing component without mod- ifying any component code or inserting additional scripts. Every standoff annotation has an annotation type identi- fier, a start position, an end position, and zero or more at- tributes. The attributes are named fields that provide infor- mation derived from or associated with the annotated text. For example, a tokenizer might emit word annotations, with text, stem and part-of-speech attributes. The start and end of each such annotation would indicate where in the text the tokenizer found the words. |
Year | Venue | Field |
|---|---|---|
2001 | TREC | Architecture,Question answering,Information retrieval,Computer science,Automation,Human–computer interaction,Artificial intelligence,Natural language processing,Language technology |
DocType | Citations | PageRank |
Conference | 6 | 1.14 |
References | Authors | |
2 | 7 | |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Pranav Anand | 1 | 260 | 19.70 |
| David Anderson | 2 | 189 | 79.84 |
| John D. Burger | 3 | 629 | 118.56 |
| John Griffith | 4 | 6 | 1.14 |
| Marc Light | 5 | 17 | 9.22 |
| Scott A. Mardis | 6 | 6 | 1.14 |
| Alexander A Morgan | 7 | 544 | 45.32 |