Update README.md

README.md

@@ -77,7 +77,7 @@ A description of all the options can be accessed as described above. The main ar
 - -th: the threshold value required for the algorithm. Default is None and, if not passed, then the programme assumes that the fit.py programme was run on some held out data to find the optimal window value automatically (see below for details on fit.py).
 - -ns: number of segments per document to be segmented. Default is None and if not provided, the programme will default to use a threshold value, assuming that the number of segments is not known. If this argument is used, pass as many integers as the number of documents to be segmented, where each integer corresponds to the number of segments in the corresponding document (in order of appearence).
 - -cfg: the location of the configuration file that can be used to run the program without the need of specifying the parameters on the command line (default: parameters.json)
-- -enc: the sentence encoder to be used. All sentence encoders from sentence_transformers library plus the DAN encoder from Universal Sentence Encoder (USE) models are available. Default is None and, if not passed in the command line, then the programme will look for it as specified in the json configuration file. In general, the names to be passed must correspond to the ones of the pretrained models from the official [sentence_transformers documentation](https://www.sbert.net/docs/pretrained_models.html) or, alternatively, pass "USE" to choose the DAN-based Universal sentence encoders released by Google on tensorflow_hub.
+- -enc: the sentence encoder to be used. All sentence encoders from sentence_transformers library plus the DAN encoder from Universal Sentence Encoder (USE) models are available. Default is None and, if not passed in the command line, then the programme will look for it as specified in the json configuration file. In general, the names to be passed must correspond to the ones of the pretrained models from the official [sentence_transformers documentation](https://www.sbert.net/docs/pretrained_models.html) or, alternatively, pass "USE" to choose the DAN-based Universal sentence encoders released by Google on [tensorflow_hub](https://tfhub.dev/google/universal-sentence-encoder/4)
 - -od: the directory where to store the results. Default is results. If any other name is passed, the programme will create the folder and, inside, will create a segments folder where to store the segmentation results and an embeddings folder where to store the computed sentence embeddings.
 - -cat: whether to concatenate all the provided documents in a single document or not. Default is False.