This task requires excellent English reading comprehension skills. Only native-quality speakers of American/British English should apply. A familiarity with Australian news is a plus. We will be hiring multiple people for this task, so once this project has been awarded, another one will be posted.
We are a university research lab working on improving how news is presented to readers. We are developing a system to track mentions of people, places, locations, brands, and other entities in the news. The system needs to be trained on manually annotated examples. This is where you come in.
You will be given an annotation tool that highlights all capitalised words and all pronouns (he, she, it, etc) in news documents. You must:
1) Confirm which highlighted sequences are the names of people, places, organisations, products, locations, etc;
2) Find any names that weren't highlighted;
3) Link together the names of the same entity within a document (e.g., you would link "Peter Dutton" to "Mr. Dutton", as they're the same person);
4) Link pronouns (he, she, we, etc) to the name they refer to (e.g. in "Peter Dutton has walked away from his seat", you would mark "his" as referring to "Peter Dutton")
5) Link the entities to Wikipedia articles.
You will be given detailed documentation of the task, and we will train you through the first few hours of the annotation. Accuracy is critical for this work, because the purpose of these documents is to train a statistical system. Incorrectly annotated documents are useless to us, because our system would learn the incorrect behaviour.
Previous annotators have averaged approximately _4 documents per hour_. Please do not assume in your bid that you will be annotating faster than this, as you probably will not be able to do so without making mistakes.
Please bid on a price to complete 60 documents, with a milestone at 10 documents. Your accuracy will be assessed over the first 10 documents to see whether we'll be able to use your annotation.
We have over 200 hours of similar annotation work to do over the next month, with more to come after that. We are looking to build relationships with reliable annotators for the future.