Semantic Representations for NLP Using VerbNet and the Generative Lexicon

Abstract

The need for deeper semantic processing of human language by our natural language processing systems is evidenced by their still-unreliable performance on inferencing tasks, even using deep learning techniques. These tasks require the detection of subtle interactions between participants in events, of sequencing of subevents that are often not explicitly mentioned, and of changes to various participants across an event. Human beings can perform this detection even when sparse lexical items are involved, suggesting that linguistic insights into these abilities could improve NLP performance. In this article, we describe new, hand-crafted semantic representations for the lexical resource VerbNet that draw heavily on the linguistic theories about subevent semantics in the Generative Lexicon (GL). VerbNet defines classes of verbs based on both their semantic and syntactic similarities, paying particular attention to shared diathesis alternations. For each class of verbs, VerbNet provides common semantic roles and typical syntactic patterns. For each syntactic pattern in a class, VerbNet defines a detailed semantic representation that traces the event participants from their initial states, through any changes and into their resulting states. The Generative Lexicon guided the structure of these representations. In GL, event structure has been integrated with dynamic semantic models in order to represent the attribute modified in the course of the event (the location of the moving entity, the extent of a created or destroyed entity, etc.) as a sequence of states related to time points or intervals. We applied that model to VerbNet semantic representations, using a class's semantic roles and a set of predicates defined across classes as components in each subevent. We will describe in detail the structure of these representations, the underlying theory that guides them, and the definition and use of the predicates. We will also evaluate the effectiveness of this resource for NLP by reviewing efforts to use the semantic representations in NLP tasks.

CS-KG: A Large-Scale Knowledge Graph of Research Entities and Claims in Computer Science

“…We address this issue by mapping similar verbs to the same predicate. The mapping schema has been built by enriching our previous handcrafted mapping [15] with VerbNet [38], which offers a complete and coherent semantic representations of verbs [10]. Verbnet is a taxonomy of English verbs organized in classes whose verbs share syntactic and semantic coherence.…”

Componential Analysis of English Verbs

“…Each VerbNet class contains a set of member verbs, the thematic roles for the predicate-argument structure of these verbs, the selectional preferences for these class-specific roles, as well as a set of typical syntactic patterns and corresponding semantic representations. These semantic representations use a Davidsonian first-order-logic formulation to provide an abstract, language-independent conceptual representation of actions, such as changes of state, changes of location, and exertion of force (Brown et al, 2019 , 2022 ). These representations use basic predicates to show the relationships between the thematic role arguments and to track any changes over the time course of the event.…”

“…For example, in the Calibratible_cos-45.6.1 class, the semantic representation traces the change of the PATIENT along a scale, with V_DIRECTION as a place-holder for the direction given as a feature for a specific verb in the class. To see more details and examples of these changes, please see Brown et al ( 2018 ), Brown et al ( 2019 ), and Brown et al ( 2022 ).…”

“…This componential view of lexical semantics, particularly as applied to the domain of verbs (Hartshorne et al, 2013 , 2014 ) has been utilized in the semantic representation of English verbs in the VerbNet lexical resource (Schuler, 2005 ; Brown et al, 2018 , 2019 , 2022 ), in the sense that English verbs have been assigned to different classes based on their semantic (and syntactic) differences, which entails that all the verbs assigned to the same class share some aspects of their semantics. Under development since 1998 (Dang et al, 1998 ; Schuler, 2005 ), VerbNet has richer and more detailed representations than PropBank (Kingsbury and Palmer, 2002 ; Palmer et al, 2005 ) and its coverage of verbs is more extensive than that of FrameNet (Baker et al, 1998 ).…”

“…It is worth noting that one of the factors that distinguishes VerbNet from other similar resources is the manually curated semantic representations for verb classes in the form of logical predicates 1 , which are shared among all the verb members of the class. In its most recent version (Brown et al, 2018 , 2019 , 2022 ), VerbNet has adopted the dynamic event structure ideas introduced by the Generative Lexicon theory (Pustejovsky, 1998 , 2013 ; Pustejovsky and Moszkowicz, 2011 ). These recent changes have also been aimed at facilitating the use of VerbNet by AI planners (McDonald et al, 2018 ).…”