Unsupervised Semantic Parsing

By Hoifung Poon. 2009

316 citations recorded by the source corpus.

We present the first unsupervised approach to the problem of learning a semantic parser, using Markov logic. Our USP system transforms dependency trees into quasi-logical forms, recursively induces lambda forms from these, and clusters them to abstract away syntactic variations of the same meaning. The MAP semantic parse of a sentence is obtained by recursively assigning its parts to lambda-form clusters and composing them. We evaluate our approach by using it to extract a knowledge base from biomedical abstracts and answer questions. USP substantially outperforms TextRunner, DIRT and an informed baseline on both precision and recall on this task.

Subfields: Artificial Intelligence, Molecular Biology, Natural Language Processing

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 Molecular Biology

In Natural Language Processing

Citations in the corpus

Learning to Solve Arithmetic Word Problems with Verb Categorization (2014, background): “More recent work (Eisenstein et al., 2009; Kate and Mooney, 2007; Goldwasser and Roth, 2011; Poon and Domingos, 2009; Goldwasser et al., 2011; Kushman and Barzilay, 2013) reduce the amount of required supervision in mapping sentences to meaning representations while taking advantage of special…”
Learning Dependency-Based Compositional Semantics (2011, background): “On the other hand, existing unsupervised semantic parsers (Poon and Domingos, 2009) do not handle deeper linguistic phenomena such as quantification, negation, and superlatives.”
Joint Video and Text Parsing for Understanding Events and Answering Queries (2013, background): “Our work also involves semantic parsing of text descriptions, which is related to a variety of semantic parsing approaches studied in the natural language processing community, for example, a supervised approach [35], an unsupervised approach [36], and a dependency based approach [37].”
Confidence Driven Unsupervised Semantic Parsing (2011, methodology): “Other approaches to unsupervised semantic analysis (Poon and Domingos, 2009; Titov and Kle-mentiev, 2011) take a different approach to semantic representation, by clustering semantically equivalent dependency tree fragments, and identifying their predicate-argument structure.”
Multi-Resolution Language Grounding with Weak Supervision (2014, background): “A few unsupervised approaches exist (Poon and Domingos, 2009; Poon, 2013), but these are specific to translating language into queries in highly structured database and cannot be applied to our more flexible domain.”
Interactive and interpretable machine learning models for human machine collaboration (2015, methodology): “MLN forms the joint distribution of a probabilistic graphical model by weighting formulas in a first-order logic [116, 121, 109, 113].”
Large-scale Semantic Parsing via Schema Matching and Lexicon Extension (2013, methodology): “More recent examples of similar techniques include the Resolver system (Yates and Et-zioni, 2009) and Poon and Domingos’s USP sys-tem (Poon and Domingos, 2009).”
Probabilistic Frame Induction (2013, methodology): “1 USP generates the argument types along with events from clustering.”
A Survey on Semantic Parsing (2018, methodology): “Unsupervised: [Poon and Domingos, 2009] was the first attempt at learning a semantic parsing model in an unsupervised manner.”
StructVAE: Tree-structured Latent Variable Models for Semi-supervised Semantic Parsing (2018, background): “There have also been efforts in unsupervised semantic parsing, which exploits external linguistic analysis of utterances (e.g., dependency trees) and the schema of target knowledge bases to infer the latent MRs (Poon and Domingos, 2009; Poon, 2013).”
From Natural Language Specifications to Program Input Parsers (2013, background): “Examples of meaning representations considered in prior research include logical forms based on database query (Tang and Mooney, 2000; Zettlemoyer and Collins, 2005; Kate and Mooney, 2007; Wong and Mooney, 2007; Poon and Domingos, 2009; Liang et al., 2011; Goldwasser et al., 2011), semantic frames (”
Annotated Gigaword (2012, methodology): “Ongoing work uses this resource to automatically extract relations, in the spirit of Lin and Pantel (2001) (DIRT) and Poon and Domingos (2009) (USP).”
Joint Learning of Words and Meaning Representations for Open-Text Semantic Parsing (2012, methodology): “wards learning to associate a MR to any kind of natural language sentence (Shi and Mihalcea, 2004; Poon and Domingos, 2009).”
Lexical Generalization in CCG Grammar Induction for Semantic Parsing (2011, background): “Other work has learnt semantic analyses from text in the context of interactions in computational environments (Branavan et al. (2010), Vogel and Juraf-sky (2010)); text grounded in partial observations of a world state (Liang et al., 2009); and from raw text alone (Poon and Domingos, 2009, 2010).”
Formalizing Visualization Design Knowledge as Constraints: Actionable and Extensible Models in Draco (2018, background): “Constraint programming is a declarative programming paradigm with wide applications in scheduling [12], graphical design [5], and natural language specification [45].”
Weakly Supervised Training of Semantic Parsers (2012, background): “Other approaches are completely unsupervised, but do not tie the language to an existing meaning representation (Poon and Domingos, 2009).”
Using Semantic Unification to Generate Regular Expressions from Natural Language (2013, background): “It has only two primitive types, integer (I) and regexp (R), with most arguments being of type R.”