Sentence BERT models trained on STS benchmark

<h3>Sentence BERT models trained on STS benchmark </h3> <p><br></p> <p>See sentence transformers (https://www.sbert.net/) for training and use of sentence BERT models in general.</p> <p>The STS benchmark dataset was fetched from https://sbert.net/datasets/stsbenchmark.tsv.gz</p> <p><br></p> <p>Models follow the naming scheme: </p> <p>sts_bert_[base model name]_[distance function]_[placement of most dissimilar]_z_[z score normalisation]_n_[normalisation of embeddings]_c_[centering of embeddings]_seed[random seed] </p> <p>Example: sts_bert_distilroberta-base_cos_dist_ORTHOGONAL_z_False_n_False_c_False_seed1</p> <p><br></p> <h4>Base model name</h4> <p>Three different base models were used:</p> <p>- distilroberta-base (https://huggingface.co/distilroberta-base)</p> <p>- microsoft-MiniLM-L12-H384-uncased (https://huggingface.co/microsoft/MiniLM-L12-H384-uncased)</p> <p>- microsoft-mpnet-base (https://huggingface.co/microsoft/mpnet-base)</p> <h4><br></h4> <h4>Distance function</h4> <p>Three different distance functions were used. </p> <p>- Cosine similarity (cos)</p> <p>- Cosine distance (cos_dist)</p> <p>- Euclidean distance (euclidean)</p> <h4><br></h4> <h4>Placement of most dissimilar</h4> <p>This describes how we want the embeddings of two texts which are completetly dissimilar to ideally be placed. </p> <p>In all these models we have chosen ORTHOGONAL, so the two embeddings of dissimilar texts should have 90 degrees between them. </p> <p>Another choice could have been OPPOSITE where the embeddings should have 180 degrees between them when they are completely dissimilar. </p> <h4><br></h4> <h4>Z-score normalisation</h4> <p>Whether z-score normalisation of the embeddings was performed during training. True/False.</p> <h4><br></h4> <h4>Normalisation of embeddings</h4> <p>Whether normalisation of embeddings was performed during training. True/False.</p> <h4><br></h4> <h4>Centering of embeddings</h4> <p>Whether centering of the embeddings was performed during training. True/False.</p> <h4><br></h4> <h4>Random seed</h4> <p>What random seed was used for training. </p> <p><br></p> <p><br></p> <p><br></p>