Measuring the Semantic Similarity of Texts

By Courtney Corley, Rada Mihalcea. EMSEE@ACL, 2005

409 citations recorded by the source corpus.

This paper presents a knowledge-based method for measuring the semantic-similarity of texts. While there is a large body of previous work focused on finding the semantic similarity of concepts and words, the application of these word-oriented methods to text similarity has not been yet explored. In this paper, we introduce a method that combines word-to-word similarity metrics into a text-to-text metric, and we show that this method outperforms the traditional text similarity metrics based on lexical matching.

Subfields: Artificial Intelligence

View Paper: https://www.semanticscholar.org/paper/6fc4a08abd9f52013ad1c598aadc7ef60aa0d7d2

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

Citations in the corpus

Paraphrase Identification as Probabilistic Quasi-Synchronous Recognition (2009, methodology): “…give the best-published classiﬁ-cation accuracy for the paraphrase identiﬁcation task (Zhang and Patrick, 2005; Finch et al., 2005; Wan et al., 2006; Corley and Mihalcea, 2005, inter alia ), but do not explicitly model correspondence structure (or “alignment”) between the parts of two sentences.”
Automatically Constructing a Corpus of Sentential Paraphrases (2005, methodology): “It has already been applied in this way by Corley & Mihalcea (2005) and Wu (2005).”
Soft Cardinality: A Parameterized Similarity Function for Text Comparison (2012, background): “…in metric spaces (e.g. cosine tf-idf similarity (Salton et al., 1975)); iii) the edit distance family of measures (e.g. Levenstein (1966) distance, LCS (Hirschberg, 1977)); and iv) hybrid approaches ((Monge and Elkan, 1996; Co-hen et al., 2003; Corley and Mihalcea, 2005; Jimenez et al., 2010)).”
Recognizing Textual Entailment Using Sentence Similarity based on Dependency Tree Skeletons (2007, methodology): “The first backup strategy is a straightforward BoW method that we will not present in this paper (see more details in (Corley and Mihalcea, 2005));”
Semantic text similarity using corpus-based word similarity and string similarity (2008, methodology): “For the paraphrase recognition task, our proposed STS method performs similarly to the combined unsupervised method [Mihalcea et al. 2006] and the combined supervised method [Corley and Mihalcea 2005].”
Have You Lost the Thread? Discovering Ongoing Conversations in Scattered Dialog Blocks (2017, background): “AsNLIis a difficult problem, many classification strategies have been proposed: (1) learning models based on lexical features (Corley and Mihalcea 2005), on syntactic features (Kouylekov and Magnini 2005; Snow et al. 2006; Wang and Neumann 2007), on multiple linguistic levels (de Salvo Braz et al.…”
Generating Phrasal and Sentential Paraphrases: A Survey of Data-Driven Methods (2010, background): “A problem closely related to, and as important as, generating paraphrases is that of assigning a quantitative measurement to the semantic similarity of two phrases (Fujita and Sato 2008a) or even two given pieces of text (Corley and Mihalcea 2005; Uzuner and Katz 2005).”
Neobility at SemEval-2017 Task 1: An Attention-based Sentence Similarity Model (2017, background): “, 2015), paraphrase detection (Fernando and Stevenson, 2008), and machine translation (Corley and Mihalcea, 2005).”
A Survey of Paraphrasing and Textual Entailment Methods (2009, background): “The greater this difference the more knowledge is required in B for (φT ∧B) |= φH to hold, and the more difficult it becomes for speakers to accept that T textually entails H. Similar bidirectional logical entailment checks can be used to recognize paraphrases (Rinaldi et al., 2003).”
Evaluating WordNet-based Measures of Lexical Semantic Relatedness (2006, background): “On the other hand, Corley and Mihalcea (2005) found little difference between the measures when using them in an algorithm for computing text similarity.”
The Third PASCAL Recognizing Textual Entailment Challenge (2007, methodology): “Interestingly, a non-participating system that operated solely at this level, using a simple decision tree trained on the development set, obtained an accuracy level of 58%, which might reflect a knowledge-poor baseline (see [5]).”
The PASCAL Recognising Textual Entailment Challenge (2005, methodology): “Interestingly, a non-participating system that operated solely at this level, using a simple decision tree trained on the development set, obtained an accuracy level of 58%, which might reflect a knowledge-poor baseline (see [5]).”
Using Dependency-Based Features to Take the ’Para-farce’ out of Paraphrase (2006, background): “Corley and Mihalcea (2005) use word equivalence features resulting in a classification accuracy of 71.5%. Zhang and Patrick (2005) examine string edit distance features and ngram overlap features collected on pairs of sentences in their canonical form.”
A Comparison of Greedy and Optimal Assessment of Natural Language Student Input Using Word-to-Word Similarity Metrics (2012, methodology): “The methods proposed so far that rely on the principle of compositionality to compute the semantic similarity of longer texts have been primarily greedy methods (Corley & Mihalcea, 2005; Lintean & Rus, 2012).”
Structured Lexical Similarity via Convolution Kernels on Dependency Trees (2011, background): “Given two text fragments d 1 and d 2 ∈ D (the text fragment set), a general lexical kernel (Basili et al., 2005) deﬁnes their similarity as: where ω 1 and ω 2 are the weights of the words (features) w 1 and w 2 in the documents d 1 and d 2 , respectively, and σ is a term similarity function, e.g. (P”
UNAL-NLP: Combining Soft Cardinality Features for Semantic Textual Similarity, Relatedness and Entailment (2014, methodology): “This set of features is motivated by the idea proposed by Corley and Mihalcea (2005) of grouping words by their POS category before being compared for semantic textual similarity.”