id	author	title	date	pages	extension	mime	words	sentences	flesch	summary	cache	txt
work_ovcflywu7jf77iwsxybgfrw4tm	Daniel Fried	Higher-order Lexical Semantic Models for Non-factoid Answer Reranking	2015	14	.pdf	application/pdf	9196	745	59	Higher-order Lexical Semantic Models for Non-factoid Answer Reranking We introduce a higher-order formalism that allows all these lexical semantic models to the semantic similarity between question and answer using language models acquired from relevant texts (Yih et al., 2013; Jansen et al., 2014). model graphs, we observe that semantic association between two words (or structures) stored in representations, including alignment and language models, over both words and syntactic structures, can be adapted to the proposed our higher-order LS models on a community question answering (CQA) task (Wang of the alignment model in §4.2, where the representations of questions and answers are changed addition, we experimented with the opposite hybrid model: interpolating the NNLM vectors using alignment associate probabilities as weights, NNLM Corpus: We generated vector representations for words using the word2vec model number of most-similar neighbor vectors interpolated when constructing a higher-order model. higher-order NNLMs and alignment models are higher-order NNLMs and alignment models are	./cache/work_ovcflywu7jf77iwsxybgfrw4tm.pdf	./txt/work_ovcflywu7jf77iwsxybgfrw4tm.txt