id	author	title	date	pages	extension	mime	words	sentences	flesch	summary	cache	txt
work_xwij57ezufevpgkcs2vwd6r76u	Sanjeev Arora	Linear Algebraic Structure of Word Senses, with Applications to Polysemy	2018	14	.pdf	application/pdf	8482	823	71	Linear Algebraic Structure of Word Senses, with Applications to Polysemy Classical vector space models (see the survey by Turney and Pantel (2010)) use simple linear algebra of embeddings, with the internal information extracted only via inner product, is felt to fail in capturing word senses (Griffiths et al., 2007; Reisinger paper strongly suggest that word embeddings computed using modern techniques such as GloVe and To do so we rely on the generative model of Section 2.1 according to which unit vector in the embedding space corresponds to a micro-topic or discourse. Table 2: Fitting the GloVe word vectors with average discourse vectors using a linear transformation. We tested three standard word embedding methods: GloVe, the skipgram variant of word2vec, and SN (Arora et al., linear algebraic structure of word senses within existing embeddings, it is desirable to verify that this	./cache/work_xwij57ezufevpgkcs2vwd6r76u.pdf	./txt/work_xwij57ezufevpgkcs2vwd6r76u.txt