Short Text Similarity with Word Embeddings

By Tom Kenter, Maarten de Rijke. 2015

485 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Natural Language Processing, Representation & Self-Supervised Learning

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Natural Language Processing

In Representation & Self-Supervised Learning

Citations in the corpus

A Deep Relevance Matching Model for Ad-hoc Retrieval (2016, methodology): “, some rare terms or numbers not trained by CBOW) in queries, we follow the practice in previous work [14] to only allow exact matching between such query terms and docu-”
Webly Supervised Joint Embedding for Cross-Modal Image-Text Retrieval (2018, background): “Averaged word vectors has been shown to be a strong feature for text in several tasks [26, 27, 58].”
Relevance-based Word Embedding (2017, background): “nd matrix factorization in [21]. „ese models have been demonstrated to be e‡ective in a number of IR tasks, including query expansion [11, 17, 40], query classi•cation [23, 41], short text similarity [15], and document model estimation [2, 31]. „e aforementioned embedding models are typically traine”
Representation learning for very short texts using weighted word embedding aggregation (2016, background): “Kenter and de Rijke (2015) learn semantic features for every sentence in the dataset based on a saliency weighted network for which the BM25 algorithm is used.”
Siamese CBOW: Optimizing Word Embeddings for Sentence Representations (2016, positive): “Surprisingly, simply averaging word embeddings of all words in a text has proven to be a strong baseline or feature across a multitude of tasks (Faruqui et al., 2014; Yu et al., 2014; Gershman and Tenenbaum, 2015; Kenter and de Rijke, 2015).”
Explaining Credibility in News Articles using Cross-Referencing (2019, methodology): “Future implementations should consider employing more sophisticated short-text similarity methods such as those based on word and sentence embeddings [6, 11].”
Retrieving Textual Evidence for Knowledge Graph Facts (2019, methodology): “Kenter and de Rijke [14] use Relaxed WMD and BM25 based weighting function as a feature in a sentence paraphrase classifier.”
Topic Modeling for Short Texts with Auxiliary Word Embeddings (2016, unknown): “Effective learning of general word semantic relatedness is now feasible and practical with the recent development in neural network techniques, which have contributed improvements in many tasks in Information Retrieval (IR) and Natural Language Processing (NLP) [9,10,16,17, 43].”
BioWordVec, improving biomedical word embeddings with subword information and MeSH (2019, methodology): “Word embeddings are often used to calculate sentence pair similarity 22 .”
Alignment over Heterogeneous Embeddings for Question Answering (2019, background): “Unsupervised alignment models have been proposed for several NLP tasks such as short text similarity (Kenter and De Rijke, 2015), answer phrase/sentence selection in reading comprehension (RC) (Chakravarti et al., 2017), document retrieval (Kim et al., 2017), etc.”
Estimating Embedding Vectors for Queries (2016, background): “Kenter and de Rijke [20] proposed a supervised approach to compute semantic similarity between short texts.”
Embedding-based Query Language Models (2016, methodology): “, the cosine similarity or the Euclidean distance functions, that are shown to be very effective in detecting nearest words in terms of their semantic similarity [12, 17].”
Sanity Check: A Strong Alignment and Information Retrieval Baseline for Question Answering (2018, unknown): “Several IR systems have been modified to use distributional similarity to align query terms to the most similar document term for various tasks, including document matching [16], short text similarity [12], and answer selection [4].”
Comprehensive stereotype content dictionaries using a semi‐automated method (2020, methodology): “If we then take two of these vectors representing two words in our dictionaries, we can measure their semantic similarity by calculating their cosine similarity (e.g., Kenter & De Rijke, 2015).”
Neural information retrieval: at the end of the early years (2017, methodology): “For instance, in the context of CNNs for text classification, Kim (2014) adopted this approach.”
Semantic Matching by Non-Linear Word Transportation for Information Retrieval (2016, methodology): “Therefore, following the practice in previous work [16], we map all the OOV words to random vectors, while remembering which OOV word maps to which random vector.”