Mining Paraphrases from Self-anchored Web Sentence Fragments

Abstract

Abstract. Near-synonyms or paraphrases are beneficial in a variety of natural language and information retrieval applications, but so far their acquisition has been confined to clean, trustworthy collections of documents with explicit external attributes. When such attributes are available, such as similar time stamps associated to a pair of news articles, previous approaches rely on them as signals of potentially high content overlap between the articles, often embodied in sentences that are only slight, paraphrase-based variations of each other. This paper introduces a new unsupervised method for extracting paraphrases from an information source of completely different nature and scale, namely unstructured text across arbitrary Web textual documents. In this case, no useful external attributes are consistently available for all documents. Instead, the paper introduces linguistically-motivated text anchors, which are identified automatically within the documents. The anchors are instrumental in the derivation of paraphrases through lightweight pairwise alignment of Web sentence fragments. A large set of categorized names, acquired separately from Web documents, serves as a filtering mechanism for improving the quality of the paraphrases. A set of paraphrases extracted from about a billion Web documents is evaluated both manually and through its impact on a natural-language Web search application.

Unsupervised Phrasal Near-Synonym Generation from Text Corpora

“…Ganitkevitch et al (2013) use the bilingual pivoting technique (Bannard and Callison-Burch 2005) along with distributional similarity features to extract lexical, and phrasal paraphrases. Some other approaches (Paşca 2005;Lin and Pantel 2001;Berant et al 2012) differ from ours in that, they use manually coded linguistic patterns to align only specific text fragment contexts to generate paraphrases (Paşca 2005), and require language specific resources such as part-of-speech taggers (Paşca 2005) and parsers (Lin and Pantel 2001). Furthermore, the latter two only find alternate constructions with the same content words, such as "X manufactures Y" infers "X's Y factory" (Lin and Pantel 2001).…”

“…The Near Synonym System (NeSS) introduces a new method which differs from other approaches in that it does not require parallel resources, (unlike Barzilay and McKeown 2001;Lin et al 2003;Callison-Burch et al 2006;Ganitkevitch et al 2013) nor does it use pre-determined sets of manually coded patterns (Lin et al 2003;Paşca, 2005). NeSS captures semantic similarity via n-gram distributional methods that implicitly preserve local syntactic structure without parsing, making the underlying method language independent.…”

Text Relatedness Based on a Word Thesaurus

“…Performance of applications relying on natural language processing may suffer from the fact that the processed documents might contain lexically different, yet semantically related, text segments. The task of recognizing synonym text segments, which is better known as paraphrase recognition, or detection, is challenging and difficult to solve, as shown in the work of Pasca (2005). The task itself is important for many text related applications, like summarization (Hirao, Fukusima, Oku-mura, Nobata, & Nanba, 2005), information extraction (Shinyama & Sekine, 2003) and question answering (Pasca, 2003).…”

Word sense disambiguation and text relatedness based on word thesauri

“…The performance of document processing applications relying on natural language processing may suffer from the fact that the processed documents might contain lexically different, yet semantically related, text segments. The task of recognizing pairs of text segments, with identical or almost identical semantics, which is better known as paraphrase detection, is challenging and difficult to solve, as shown in the work of Mihalcea et al [66], and Pasca [85]. The task itself is important for many text related applications, like summarization [38], information extraction [104] and question answering [84].…”