Training data. In order to accurately identify query-topic, we need training data representative of the topic-classes. We use query-based sampling techniques for obtaining topic-descriptions, similar to the sampling described in the Section 4.1. We used the same set of ▇▇▇- pling methods and list of queries described in Sec- tion 4. 1. But instead of generating partial queries by deleting words randomly, full titles or names are used as queries.
Appears in 2 contracts
Sources: Source Selection Agreement, Source Selection Agreement