Table 6.
Reference resolution features
| Feature Name | Feature Types | Description |
|---|---|---|
| closestTempDist | other | The distance of the closest template in a candidate cluster to the current template |
| containedIn | anatomic | If any of the anatomies in the current template are contained in the anatomy in the candidate cluster |
| containerOf | anatomic | If any of the anatomies in the candidate cluster are contained in the current template |
| header | static | If the sentence of the template looks like a section header |
| isSuperset | other | If the candidate cluster is already a superset of another cluster |
| malignancy | static | Malignancy status of template |
| malignancyOfCandCluster | static | Malignancy status of the cluster |
| nextBestSim | relative, similarity | L-2 norm of the next best similarity vector |
| ngrams | static | 1-, 2-, and 3- grams (using lemma) for sentences of template and a candidate cluster |
| ngramsMatching | other | Matching 1-, 2-, and 3- grams (using raw words) for sentences of template and a candidate cluster |
| nthTemplate | positional, static | The number template in the document |
| numOfCand | other | The number of candidate clusters |
| numOfMeas | static | The number of measurements |
| numOfTempInCluster | other | The number of templates in the candidate cluster |
| onlySameMal | relative | The only candidate cluster with matching malignancy as template |
| onlySameMeas | relative | The only candidate cluster with matching measurement malignancy as template |
| sameOrgan | anatomic | If the organ in the sentence matches organ in a cluster |
| sameLocations | anatomic | The matching locations of all |
| section | static | Section of the template |
| sim | similarity | The L2-norm of similarity vector |
| simvecfeats | similarity | This feature extends from the similarity vector features so that each individual similarity vector dimensions are each considered their own feature |
| summaryOf | static | If tumor reference is preceded with “the”, “this”, “these” |
| totalNumOfTemp | static | Total number of templates in the document |
| totalNumOfImpTemp | static | Total number of templates in the Impressions section |
| UMLS | other | Matching UMLS concept between the template and the cluster |
| Underheading | other | If there is a sentence belonging in the cluster that looks like a header of the current template |