Alastair Butler

DynamicPower at SemEval-2016 Task 8: Processing syntactic parse trees with a Dynamic Semantics core.

This is a system description paper for a submission to Task 8 of SemEval-2016: Meaning Representation Parsing. No use was made of the training data provided by the task. Instead existing components were combined to form a pipeline able to take raw sentences as input and output meaning representations. Components are a part-of-speech tagger and parser trained on the Penn Parsed Corpus of Modern British English to produce syntactic parse trees, a semantic role labeller and a named entity recogniser to supplement obtained parse trees with word sense, functional and named entity information, followed by an adapted Tarskian satisfaction relation for a Dynamic Semantics that is used to transform a syntactic parse into a predicate logic based meaning representation, followed by conversion to penman/AMR notation required for the task appraisal.

Annotating Syntax and Lexical Semantics With(out) Indexing

A new method of annotation is introduced that is able to establish a rich range of dependencies without recourse to indexing. With well defined annotation practices for assigning structure, the index-less annotation can be subsequently processed by means of a mechanism of semantic calculation that identifies relationships (dependencies) by reference to structure. These relationships can be subsequently expressed through a derived indexing, but don’t rely on indexing in order to be established. The technique is illustrated capturing dependencies seen with PropBank/FrameNet annotations.

Deterministic natural language generation from meaning representations for machine translation.

This paper describes a deterministic method for generating natural language suited to being part of a machine translation system with meaning representations as the level for language transfer. Starting from Davidsonian/Penman meaning representations, syntactic trees are built following the Penn Parsed Corpus of Modern British English, from which the yield (i.e., the words) can be taken. The novel contribution is to highlight exploiting the presentation of meaning content to inform decisions regarding the selection of language constructions: active vs. passive, argument subject vs. expletive it vs. existential there, discourse vs. intra-sentential coordination vs. adverbial clause vs. participial clause vs. purpose clause, and infinitive clause vs. finite clause vs. small clause vs. relative clause vs. it cleft.

Round trips with meaning stopovers

This paper describes taking parsed sentences, going to meaning representations (the stopover), and then back to parsed sentences (the round trip). Keeping to the same language tests the combined success of building meaning representations from parsed input and of generating parsed output. Switching languages when manipulating meaning representations would achieve translation. Transfer shortfall is seen with meaning representations built from parsed parallel corpora data, with English-Japanese as an example.

Self-Locating Evaluation

This chapter develops utilising the assignment as a source of information about the content of the expression under evaluation, with assignments governing where dependencies are located throughout an evaluation. This is achieved with a language that includes fine grained and often inter dependent primitive operations of scope manipulation to support processing structures close to expected conventional parsings of natural language.

Self-Selective Evaluation

This chapter adds an If conditional to a recursive routine for evaluating expressions against an assignment function that stores accumulated binding information. Assignments are considered that assign sequences as values, so what is evaluated can be selected based on tests regarding lengths of assigned sequences. This allows for a robust interpretation of unknown lexical items and for feeding an automated regulation of binding information to leave little need for explicitly coding dependencies.

Coordinating and Subordinating Binding Dependencies

This chapter focuses on similarities between coordinating and (distant) subordinating binding dependencies. We start from natural language data suggesting such dependencies are established by the same underlying mechanism, e.g., “A collector didn’t buy because she was influenced.” is structurally ambiguous without consequences for the pronominal binding. We compare and contrast four related systems that capture coordinating and subordinating binding dependencies, the first with distinct mechanisms, the others with single mechanisms.

Treebank annotation for formal semantics research

This paper motivates and describes treebank annotation for Japanese and English following a scheme adapted from the Annotation manual for the Penn Historical Corpora and the PCEEC (Santorini 2010). The purpose of this annotation is to create a syntactic base from which meaning representations can be built automatically on a corpus linguistics scale (thousands of examples). Advantages of the adopted annotation scheme are highlighted. Most notably, marking clause level functional information is essential for deterministically building meaning representations beyond the predicate-argument structure level. Also an internal syntax where phrasal categories are fundamentally similar is of great assistance. Finally, the paper demonstrates how scope information is simple to add when bracketed syntactic structure is inherently flat.

Logic and Engineering of Natural Language Semantics (LENLS) 8

On December 1-2, 2011 the Eighth International Workshop of Logic and Engineering of Natural Language Semantics (LENLS 8) took place at Sunport Hall Takamatsu. This was held as a workshop of the third JSAI International Symposia on AI (JSAI-isAI 2011), sponsored by The Japan Society for Artificial Intelligence (JSAI).

Towards a Self-selective and Self-healing Evaluation

Assume a recursive routine for evaluating expressions against an assignment function that stores accumulated binding information for variable names. This paper proposes adding an If operation that allows for what is evaluated to be automatically selected during the runtime of evaluation based on the state of the assignment function. This can (a) allow a single encoding of content that would otherwise require distinct expressions, and (b) equip an expression with a way to recover from situations that would cause unwelcome results from evaluation. The new operation is demonstrated to be an essential component for allowing a robust interpretation of unknown lexical items and for feeding an automated regulation of binding dependencies determined on a grammatical basis.