Contract Discovery: Dataset and a Few-Shot Semantic Retrieval Challenge with Competitive Baselines

Benchmark Model Rank Results
semantic-retrieval-on-contract-discoveryHuman baseline#1Soft-F1: 0.84
semantic-retrieval-on-contract-discoveryk-NN with sentence n-grams, GPT-2 embeddings, fICA#2Soft-F1: 0.51
semantic-retrieval-on-contract-discoveryLSA baseline#3Soft-F1: 0.39
semantic-retrieval-on-contract-discoveryUniversal Sentence Encoder#4Soft-F1: 0.38
semantic-retrieval-on-contract-discoverySentence BERT#5Soft-F1: 0.31