ShotgunWSD: An unsupervised algorithm for global word sense disambiguation inspired by DNA sequencing

Abstract

In this paper, we present a novel unsupervised algorithm for word sense disambiguation (WSD) at the document level. Our algorithm is inspired by a widely-used approach in the field of genetics for whole genome sequencing, known as the Shotgun sequencing technique. The proposed WSD algorithm is based on three main steps. First, a brute-force WSD algorithm is applied to short context windows (up to 10 words) selected from the document in order to generate a short list of likely sense configurations for each window. In the second step, these local sense configurations are assembled into longer composite configurations based on suffix and prefix matching. The resulted configurations are ranked by their length, and the sense of each word is chosen based on a voting scheme that considers only the top k configurations in which the word appears. We compare our algorithm with other state-of-the-art unsupervised WSD algorithms and demonstrate better performance, sometimes by a very large margin. We also show that our algorithm can yield better performance than the Most Common Sense (MCS) baseline on one data set. Moreover, our algorithm has a very small number of parameters, is robust to parameter tuning, and, unlike other bioinspired methods, it gives a deterministic solution (it does not involve random choices).

ShotgunWSD 2.0: An Improved Algorithm for Global Word Sense Disambiguation

“…A score is assigned to each sense configuration by computing the semantic relatedness between word senses (steps 16-19), as described by Patwardhan et al [33]. Butnaru et al [25] alternatively employed two measures to compute the semantic relatedness, one is the extended Lesk measure [31], [32] and the other is a simple approach based on deriving sense embeddings from word embeddings [36]. In this paper, we propose a third approach that is based on clustering word vectors with k-means and on eliminating the smaller clusters (which contain outlier words).…”

“…In this paper, we present an improved version of a recently introduced WSD algorithm [25], termed ShotgunWSD, 1 which stems from the Shotgun genome sequencing technique [26], [27]. ShotgunWSD is unsupervised, but it also requires knowledge in the form of WordNet synsets and relations [28], [29].…”

“…It employs a local WSD algorithm to build the local sense configurations. Butnaru et al [25] alternatively used two methods for this step, namely the extended Lesk measure [31], [32] and an approach based on deriving sense embeddings from word embeddings [34]- [36].…”

“…We present experiments on SemEval 2007 [37], Senseval-2 [38], Senseval-3 [39], SemEval 2013 [40], SemEval 2015 [41], as well as on the unified data sets [11], in order to compare ShotgunWSD 2.0 with its previous version [25], other state-of-the-art unsupervised and knowledge-based approaches [15], [18]- [23], [42]- [45], as well as the Most Common Sense (MCS) baseline. 2 MCS is considered as one of the strongest baselines in WSD [7], surpassing all unsupervised approaches in the recent SemEval 2015 [41] WSD task.…”

UnibucKernel: A kernel-based learning method for complex word identification

“…Based on WordNet, we form the sense bag for a given synset by collecting the words found in the gloss of the synset (examples included) as well as the words found in the glosses of semantically related synsets. The semantic relations are chosen based on the part-ofspeech of the target word, as described in (Butnaru et al, 2017). To derive the sense embedding, we embed the collected words in an embedding space and compute the median of the resulted word vectors.…”