Minimal Recursion Semantics Research Articles

Generation of MRS Abstract Predicates from Paninian USR

Semantic Representations become useful resources for various multilingual NLP applications such as Machine Translation, Multilingual Generation, cross Lingual QA, to name a few. No Semantic Representation, to our knowledge, adopts vivakṣā (Speaker’s intention) as a guiding principle for the representation. This motivates us to develop a new Semantic Representation system – Universal Semantic Representation (USR) – following Indian Grammatical Tradition (IGT) and Paninian grammar. Since USR is designed to be language-independent, we have currently taken up the task of generating English, Hindi, Tamil and Bangla from the USR. For English generation, the USR is mapped to ERG meaning representation (Flickinger, D. 1999) which is couched in Minimal Recursion Semantics (MRS). We use an off-the-shelf ACE generator that uses ERG as a resource-grammar for generating English. While designing the transfer module from USR to ERG-based MRS, we came across various Abstract Predicates (APs) in MRS representation as described in ErgSemantics_Basic (Flickinger et al., 2014). These APs are used to represent the semantic contribution of grammatical constructions or more specialized lexical entries such as compounding or the comparative use of more and so on. This paper presents the strategy for postulating the APs from the information given in USR and then reports the implementation of the transfer module keeping the focus on the postulation of APs. We get around 95% accuracy in postulating APs from USR.

20 years of the Grammar Matrix: cross-linguistic hypothesis testing of increasingly complex interactions

The Grammar Matrix project is a meta-grammar engineering framework expressed in Head-driven Phrase Structure Grammar (HPSG) and Minimal Recursion Semantics (MRS). It automates grammar implementation and is thus a tool and a resource for linguistic hypothesis testing at scale. In this paper, we summarize how the Grammar Matrix grew in the last decade and describe how new additions to the system have made it possible to study interactions between analyses, both monolingually and cross-linguistically, at new levels of complexity.

HPSG analysis and computational implementation of Indonesian passives

This study aims to analyze and develop a detailed model of syntax and semantics of passive sentences in standard Indonesian in the framework of Head-Driven Phrase Structure Grammar (HPSG) (Pollard & Sag, 1994; Sag et al., 2003) and Minimal Recursion Semantics (MRS) (Copestake et al., 2005), explicit enough to be interpreted by a computer, focusing on implementation rather than theory. There are two main types of passive in Indonesian, following Sneddon et al. (2010, pp. 256-260) and Alwi et al. (2014, pp. 352-356), called ‘passive type 1’ (P1) and ‘passive type 2’ (P2). Both types were analyzed and implemented in the Indonesian Resource Grammar (INDRA), a computational grammar for Indonesian (Moeljadi et al., 2015).

Special Thematic Section on Semantic Models for Natural Language Processing (Preface)

Abstract With the availability of large language data online, cross-linked lexical resources (such as BabelNet, Predicate Matrix and UBY) and semantically annotated corpora (SemCor, OntoNotes, etc.), more and more applications in Natural Language Processing (NLP) have started to exploit various semantic models. The semantic models have been created on the base of LSA, clustering, word embeddings, deep learning, neural networks, etc., and abstract logical forms, such as Minimal Recursion Semantics (MRS) or Abstract Meaning Representation (AMR), etc. Additionally, the Linguistic Linked Open Data Cloud has been initiated (LLOD Cloud) which interlinks linguistic data for improving the tasks of NLP. This cloud has been expanding enormously for the last four-five years. It includes corpora, lexicons, thesauri, knowledge bases of various kinds, organized around appropriate ontologies, such as LEMON. The semantic models behind the data organization as well as the representation of the semantic resources themselves are a challenge to the NLP community. The NLP applications that extensively rely on the above discussed models include Machine Translation, Information Extraction, Question Answering, Text Simplification, etc.

A Notion of Semantic Coherence for Underspecified Semantic Representation

The general problem of finding satisfying solutions to constraint-based underspecified representations of quantifier scope is NP-complete. Existing frameworks, including Dominance Graphs, Minimal Recursion Semantics, and Hole Semantics, have struggled to balance expressivity and tractability in order to cover real natural language sentences with efficient algorithms. We address this trade-off with a general principle of coherence, which requires that every variable introduced in the domain of discourse must contribute to the overall semantics of the sentence. We show that every underspecified representation meeting this criterion can be efficiently processed, and that our set of representations subsumes all previously identified tractable sets.

Divergence in expressing definiteness between Mandarin and Cantonese

In this paper, we model the dialectal variation in the expression of definiteness in Mandarin and Cantonese adopting the Head-Driven Phrase Structure Grammar (HPSG) framework (Pollard & Sag, 1994) and Minimal Recursion Semantics (MRS) (Copestake et al., 2005).

A Base-generated Approach to Why-Stripping in English

Recently, Yoshida et al. (2014) propose a movement-plus-ellipsis analysis of the Why-Stripping construction with base-generation of why in the clause initial position. While agreeing with the idea of base-generated why in this construction, this paper argues that Yoshida et al. have a number of limitations, especially with respect to the clausal ellipsis requirement and the analysis of Stripping constructions with other adverbials. In this paper, a non-movement, non-ellipsis based analysis of the construction will be proposed on the basis of Ginzburg & Sag’s (2000) approach to fragment clauses. By incorporating Minimal Recursion Semantics representations, this paper also provides an account of question interpretations, focus-sensitive readings, and scope relations in why and other adverbial Stripping constructions.

Ivan A. Sag

Ivan A. Sag

Mandarin Text Representation Based on Minimal Recursion Semantics——Illustrate by Quantitative Noun Phrases

Mandarin Text Representation Based on Minimal Recursion Semantics——Illustrate by Quantitative Noun Phrases

The direct evidential -te in Korean: Its interaction with person and experiencer predicates

This paper discusses the complex relations among the direct evidential -te, person, and experiencer predicates in Korean. The questions of the paper are: (i) how the three components are related with each other in the evidential sentences, and (ii) how the interactions of the three components can be formally analyzed to correctly license only the well-formed evidential sentences. I show that in direct evidential construction with a non-private predicate (e.g. pwutulep- ‘soft’), the asserter/epistemic authority (i.e. the speaker na ‘I’ in declarative or the addressee ne ‘you’ in question) must be the experiencer of the predicate, but there is no such constraint in direct evidential construction with a private predicate (e.g. aphu- ‘sick’). I also show that the direct evidential construction with a non-private predicate is an instance of self-ascription. Then I propose an analysis of the experiencer predicates and associated lexical rules in the Minimal Recursion Semantics (MRS) (Copestake, et al., 2005) of Head-driven Phrase Structure Grammar (HPSG) (Pollard and Sag, 1994; Sag, et al., 2003).

Detecting modification of biomedical events using a deep parsing approach

BackgroundThis work describes a system for identifying event mentions in bio-molecular research abstracts that are either speculative (e.g. analysis of IkappaBalpha phosphorylation, where it is not specified whether phosphorylation did or did not occur) or negated (e.g. inhibition of IkappaBalpha phosphorylation, where phosphorylation did not occur). The data comes from a standard dataset created for the BioNLP 2009 Shared Task. The system uses a machine-learning approach, where the features used for classification are a combination of shallow features derived from the words of the sentences and more complex features based on the semantic outputs produced by a deep parser.MethodTo detect event modification, we use a Maximum Entropy learner with features extracted from the data relative to the trigger words of the events. The shallow features are bag-of-words features based on a small sliding context window of 3-4 tokens on either side of the trigger word. The deep parser features are derived from parses produced by the English Resource Grammar and the RASP parser. The outputs of these parsers are converted into the Minimal Recursion Semantics formalism, and from this, we extract features motivated by linguistics and the data itself. All of these features are combined to create training or test data for the machine learning algorithm.ResultsOver the test data, our methods produce approximately a 4% absolute increase in F-score for detection of event modification compared to a baseline based only on the shallow bag-of-words features.ConclusionsOur results indicate that grammar-based techniques can enhance the accuracy of methods for detecting event modification.

Semantics-based Question Generation and Implementation

This paper presents a question generation system based on the approach of semantic rewriting. The state-of-the-art deep linguistic parsing and generation tools are employed to convert (back and forth) between the natural language sentences and their meaning representations in the form of Minimal Recursion Semantics (MRS). By carefully operating on the semantic structures, we show a principled way of generating questions without ad-hoc manipulation of the syntactic structures. Based on the (partial) understanding of the sentence meaning, the system generates questions which are semantically grounded and purposeful. And with the support of deep linguistic grammars, the grammaticality of the generation results is warranted. Further, with a specialized ranking model, the linguistic realizations from the general purpose generation model are further refined for our the question generation task. The evaluation results from QGSTEC2010 show promising prospects of the proposed approach.

Preposed negation in Danish

In Danish the base position of the negation and negated quantifier phrases is between the subject and the finite verb in embedded clauses. However, in embedded clauses introduced by a non-veridical complementizer such as hvis (ˋif') or om (ˋwhether') the negation and negated quantifier phrases can also appear between the complementizer and the subject. This phenomenon is referred to as preposed negation. The paper investigates the structure and semantics of this construction. It is argued that preposed negation is no adjunction structure, but a special construction where the negation element is a sister of the complementizer and the filler of a filler-gap-structure. It is further argued that preposed negation is associated with negated verum-focus of a clause lacking an (aboutness-) topic. The negation of a verum predicate explains why preposed negation fails to license strong negative polarity items and to rule out positive ones. The lack of a topic explains why preposed negation is preferred with non-referential subjects and with weak readings of indefinite subjects and why preposed negation is incompatible with topic-binding particles.The final section presents an HPSG-analysis of preposed negation using Minimal Recursion Semantics (MRS).

A Minimal Recursion Semantic Analysis of Locatives

The article describes a pilot implementation of a grammar containing different types of locative PPs. In particular, we investigate the distinction between static and directional locatives, and between different types of directional locatives. Locatives may act as modifiers as well as referring expressions depending on the syntactic context. We handle this with a single lexical entry. The implementation is of Norwegian locatives, but English locatives are both discussed and compared to Norwegian locatives. The semantic analysis is based on a proposal by Markus Kracht (2002), and we show how this analysis can be incorporated into Minimal Recursion Semantics (MRS) (Copestake et al. 2005). We discuss how the resulting system may be applied in a transfer-based machine translation system, and how we can map from a shallow MRS representation to a deeper semantic representation.

Constraining aspectual composition

In Modern Greek there is a rich aspectual system, which involves both morphologically expressed grammatical aspect and eventuality types, carried primarily by the meaning of the verbal predicate. Particular emphasis is paid to the interaction between grammatical aspect and eventuality types, since it is due to this interaction that the verbal predicate acquires distinct meanings. In order to explain potential changes in the meaning of the eventualities caused by the interaction with grammatical aspect, I propose a formal analysis within HPSG, using Minimal Recursion Semantics (MRS) for the semantic representations. Following the MRS architecture, I introduce a number of relations, which represent both grammatical aspect and eventuality types. The close interaction between grammatical aspect and eventuality types triggers special meanings which traditionally can be explained by inserting contextual information into the representations. In this paper, I argue against such an analysis, providing an alternative which is based on the introduction of subeventual templates formulated by Michaelis (2003) and Pustejovsky (1995). In this context, grammatical aspect combines with eventuality types and selects eventualities or subeventualities appropriate to its selection restrictions, using information that is already there in the denotation of the eventualities.

Minimal Recursion Semantics: An Introduction

Minimal recursion semantics (MRS) is a framework for computational semantics that is suitable for parsing and generation and that can be implemented in typed feature structure formalisms. We discuss why, in general, a semantic representation with minimal structure is desirable and illustrate how a descriptively adequate representation with a nonrecursive structure may be achieved. MRS enables a simple formulation of the grammatical constraints on lexical and phrasal semantics, including the principles of semantic composition. We have integrated MRS with a broad-coverage HPSG grammar.

Between shifts and alternations: Ditransitive constructions

In this paper I present two classes of double object constructions in Modern Greek, i.e., the genitive, as well as the double accusative, ditransitive constructions. I show that these two classes differ from one another in that not both of them permit derivational processes such as the formation of adjectival passives. I also look at the case properties associated with the verbs which head Modern Greek genitive and double accusative ditransitive constructions. Finally, the analysis I propose for these constructions in Modern Greek are formalized using the Minimal Recursion Semantics (MRS) framework of Copestake et al. (2001) and Copestake et al. (2003).

A treatment of directionals in two implemented HPSG grammars

This article proposes a semantics of directional expressions in Norwegian and German, regarded as VP modifiers. The analysis uses Minimal Recursion Semantics, as an integrated part of Matrix-based HPSG grammars. Directional expressions are analyzed as modifying an individual, the 'mover'. Context dependent directionals like here receive a decomposed analysis. Telicity values reflecting various types of directional and locative expressions are computed.

Glue semantics for HPSG

‎The glue approach to semantic interpretation has been developed principally for Lexical Functional Grammar. Recent work has shown how glue can be used with a variety of syntactic theories and this paper outlines how it can be applied to HPSG. As well as providing an alternative form of semantics for HPSG, we believe that the benefits of HPSG glue include the following: (1) simplification of the Semantics Principle; (2) a simple and elegant treatment of modifier scope, including empirical phenomena like quantifier scope ambiguity, the interaction of scope with raising, and recursive modification; (3) an analysis of control that handles agreement between controlled subjects and their coarguments while allowing for a property denotation for the controlled clause; (4) re-use of highly efficient techniques for semantic derivation already implemented for LFG, and which target problems of ambiguity management also addressed by Minimal Recursion Semantics.

Syntax-semantics interface for tense and aspect in French

This paper proposes an HPSG account of the French tense and aspect system, focussing on the analysis of the passé simple (simple past) and imparfait (imperfective) tenses and their interaction with aspectually sensitive adjuncts. Starting from de Swart's (1998) analysis of the semantics of tense and aspect, I show that while the proposed semantic representations are appropriate, the analysis of implicit aspectual operators as coercion operators is inadequate. The proposed HPSG analysis relies on Minimal Recursion Semantics to relate standard syntactic structures with de Swart-style semantic representations. The analysis has two crucial features: first, it assumes that the semantic contribution of tense originates in the verb's semantic representation, despite the fact that tense can get wide scope over other semantic elements. Second, it allows the occurrence of implicit aspectual operators to be controlled by the verb's inflectional class, which accounts for their peculiar distribution.