engleski

Inferring standard name form, gender and nobility from historical texts using stable model semantics

Statistical and rule-based models were evaluated as trained on initial modern language training set and after improving models. The SM Rules model was improved by hand-writing four additional rules that were obvious from the errors in the first model. The CRF model was improved by labeling 100 addition records from the source and adding these to the dataset. This task took approximately twice the time it took to write the rules. This can be considered to represent a similar investment of resources, although this is not a precise measure of the effort invested in improving the models because both procedures depended on the characteristics of the datasets and the experience of the researcher.

Infering standard name form ; Stable model semantics

nije evidentirano