id	author	title	date	pages	extension	mime	words	sentences	flesch	summary	cache	txt
work_7jjkktfhpnadnpukptzosc5hta	Piotr Bojanowski	Enriching Word Vectors with Subword Information	2017	12	.pdf	application/pdf	6904	814	69	recently proposed morphological word representations, we show that our vectors achieve (2013b) proposed simple log-bilinear models to learn continuous representations of words on very large corpora neural language models, where words are represented as sets of features. Recently, several works have proposed different composition functions to derive representations of words Word representations trained on morphologically annotated data were introduced by Cotterell and Schütze (2015). have proposed using subword units to obtain representations of rare words (Sennrich et al., 2016; Luong and Manning, 2016). First, by looking at Table 1, we notice that the proposed model (sisg), which uses subword information, outperforms the baselines on all datasets except and obtain representations for out-of-vocabulary words with our model by summing the vectors of character of our word vectors on the similarity task as a function of the training data size. the language model with pre-trained word representations improves the test perplexity over the baseline LSTM.	./cache/work_7jjkktfhpnadnpukptzosc5hta.pdf	./txt/work_7jjkktfhpnadnpukptzosc5hta.txt