PPDB 2.0: Better paraphrase ranking, fine-grained entailment relations, word embeddings, and style classification

By Ellie Pavlick, Pushpendre Rastogi, Juri Ganitkevitch, Benjamin Van Durme, Chris Callison-Burch. 2015

342 citations recorded by the source corpus.

We present a new release of the Paraphrase Database. PPDB 2.0 includes a discriminatively re-ranked set of paraphrases that achieve a higher correlation with human judgments than PPDB 1.0’s heuristic rankings. Each paraphrase pair in the database now also includes finegrained entailment relations, word embedding similarities, and style annotations.

Subfields: Artificial Intelligence, Natural Language Processing, Evaluation & Benchmarks

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 Evaluation & Benchmarks

Citations in the corpus

A large annotated corpus for learning natural language inference (2015, background): “Outside the domain of sentence-level entailment, Levy et al. (2014) introduce a large corpus of semi-automatically annotated entailment examples between subject–verb– object relation triples, and the second release of the Paraphrase Database (Pavlick et al., 2015) includes automatically generated en”
ProoFVer: Natural Logic Theorem Proving for Fact Verification (2021, methodology): “We assign a NatOp to individual mutations from the mutation sequence by relying on external lexicons and KBs. Specifically, we use Wordnet (Miller, 1995), Wikidata (Vrandečić and Krötzsch, 2014), Paraphrase Database (Ganitkevitch et al., 2013, PPDB), and negative sentiment polarity words from the ”
A Survey of Data Augmentation Approaches for NLP (2021, background): “Know : Short for external knowledge, this column is (cid:51) when the data augmentation process requires knowledge resources which go beyond the immediate input examples and the task deﬁnition, such as WordNet (Miller, 1995) or PPDB (Pavlick et al., 2015).”
Towards Universal Paraphrastic Sentence Embeddings (2015, methodology): “One is a large sample of 26,456 annotated phrase pairs developed by Pavlick et al. (2015).”
Paraphrasing Revisited with Neural Machine Translation (2017, methodology): “Furthermore, despite using fewer features and pivot languages, it obtains a closer correspondence to human data compared to PPDB 2.0 (Pavlick et al., 2015).”
Detecting Asymmetric Semantic Relations in Context: A Case-Study on Hypernymy Detection (2017, methodology): “It has been created us-ing 375 word pairs from a subset of the English Paraphrase Database (Ganitkevitch et al., 2013; Pavlick et al., 2015).”
Paraphrase Augmented Task-Oriented Dialog Generation (2020, background): “It re-places a sequence of tokens (text span) by their paraphrase candidates from the lexicon database (PPDB (Pavlick et al., 2015)).”
Teaching Machine Comprehension with Compositional Explanations (2020, methodology): “We tried various data as proxies, including coreference resolution data (coreference result on the SQuAD dataset produced by the Stanford CoreNLP (Manning et al., 2014)) and existing paraphrase data (PPDB (Pavlick et al., 2015)).”
PurdueNLP at SemEval-2017 Task 1: Predicting Semantic Textual Similarity with Paraphrase and Event Embeddings (2017, methodology): “We train PE using two datasets, PPDB 2.0 (Pavlick et al., 2015) and SemEval STS data. used in the ﬁrst (Section 2.1) and second steps (Section 2.2) respectively.”
Emergent: a novel data-set for stance classification (2016, methodology): “Therefore, based on the work by Rus and Lintean (2012) we compute an alignment using the Paraphrase Database (PPDB) (Pavlick et al., 2015) and the Kuhn-Munkres algo-rithm (Kuhn, 1955; Munkres, 1957) as follows.”
Unsupervised Post-processing of Word Vectors via Conceptor Negation (2018, methodology): “2016; 2017), where the linguistic constraints are extracted from an external linguistic knowledge base such as WordNet (Miller 1995) and PPDB (Pavlick et al. 2015).”
Intent detection using semantically enriched word embeddings (2016, methodology): “To deal with this issue, semantic lexicons such as WordNet [5] and the Paraphrase Database (PPDB) [6, 7], which contain semantic relations among words, have been used to enrich word embeddings.”
MUSS: Multilingual Unsupervised Sentence Simplification by Mining Paraphrases (2022, neutral): “Bilingual pivoting relied on MT datasets to create large databases of word-level paraphrases (Pavlick et al., 2015), lexical simplifications (Pavlick and Callison-Burch, 2016; Kriz et al., 2018), or sentence-level paraphrase corpora (Wieting and Gim-pel, 2018).”
A Continuously Growing Dataset of Sentential Paraphrases (2017, methodology): “Table 12: Coverage comparison of phrasal para-phrases extracted from Twitter URL data (sampled 1:1:1 from GIZA++, Jacana and Sultan’s aligner outputs) and the PPDB (Ganitkevitch et al., 2013). lated versions of a foreign novel (Barzilay and Elhadad, 2003; Barzilay and Lee, 2003) or other texts (Cohn”
Open Subtitles Paraphrase Corpus for Six Languages (2018, methodology): “Para-phrase resources have been published earlier, for instance by Quirk et al. (2004), Dolan et al. (2004), Dolan and Brockett (2005), Ganitkevitch et al. (2013), Ganitkevitch and Callison-Burch (2014), and Pavlick et al. (2015).”
DT_Team at SemEval-2017 Task 1: Semantic Similarity Using Alignments, Sentence-Level Embeddings and Gaussian Mixture Model Output (2017, methodology): “We used the word2vec (Mikolov et al., 2013) vectorial word representation, PPDB database (Pavlick et al., 2015) 2 , and WordNet (Miller, 1995) to compute similarity between words.”