Parameter Setting Sample Clauses

Parameter Setting. Our entity re-ranking approach involves free parameter λ that needs to be estimated (see Eq. (6)). To set this parameter, we employed the Coordinate Ascent algorithm [17] with random restart of 3, optimized for NDCG@100. All experiments were performed using 5-fold cross-validation, where the folds were obtained from the collection (DBpedia-Entity V2). This makes our results com- parable to the DBpedia-Entity V2 baseline runs, as the same folds are used for all the methods. Entity re-ranking was performed on top 1000 entities ranked by two state-of-the-art term-based entity retrieval models: FSDM and BM25F- CA [12]. For all experiments, we used the embedding vectors of 100 dimensions, which were trained using the settings described in Sect. 4.2.

Parameter Setting. Port: 10