Information and argument structures in Sociology research abstracts

Semantic frames are formal structures describing situations, actions or events, e.g., Commerce buy, Kidnapping, or Exchange. Each frame provides a set of frame elements or semantic roles corresponding to participants of the situation and lexical units (LUs)—words and phrases that can evoke this particular frame in texts. For example, for the frame Kidnapping, two key roles are Perpetrator and the Victim, and this frame can be evoked with lexical units abduct, kidnap, or snatcher. While formally sound, the scarce availability of semantic frame resources and their limited lexical coverage hinders the wider adoption of frame semantics across languages and domains. To tackle this problem, firstly, we propose a method that takes as input a few frame-annotated sentences and generates alternative lexical realizations of lexical units and semantic roles matching the original frame definition. Secondly, we show that the obtained synthetically generated semantic frame annotated examples help to improve the quality of frame-semantic parsing. To evaluate our proposed approach, we decompose our work into two parts. In the first part of text augmentation for LUs and roles, we experiment with various types of models such as distributional thesauri, non-contextualized word embeddings (word2vec, fastText, GloVe), and Transformer-based contextualized models, such as BERT or XLNet. We perform the intrinsic evaluation of these induced lexical substitutes using FrameNet gold annotations. Models based on Transformers show overall superior performance, however, they do not always outperform simpler models (based on static embeddings) unless information about the target word is suitably injected. However, we observe that non-contextualized models also show comparable performance on the task of LU expansion. We also show that combining substitutes of individual models can significantly improve the quality of final substitutes. Because intrinsic evaluation scores are highly dependent on the gold dataset and the frame preservation, and cannot be ensured by an automatic evaluation mechanism because of the incompleteness of gold datasets, we also carried out experiments with manual evaluation on sample datasets to further analyze the usefulness of our approach. The results show that the manual evaluation framework significantly outperforms automatic evaluation for lexical substitution. For extrinsic evaluation, the second part of this work assesses the utility of these lexical substitutes for the improvement of frame-semantic parsing. We took a small set of frame-annotated sentences and augmented them by replacing corresponding target words with their closest substitutes, obtained from best-performing models. Our extensive experiments on the original and augmented set of annotations with two semantic parsers show that our method is effective for improving the downstream parsing task by training set augmentation, as well as for quickly building FrameNet-like resources for new languages or subject domains.

In this article, we introduce a project aimed at enhancing a valency lexicon of Czech verbs with semantic information. For this purpose, we make use of FrameNet, a semantically oriented lexical resource. At the present stage, semantic frames from FrameNet have been mapped to eight groups of verbs with various semantic and syntactic properties. The feasibility of this task has been verified by the achieved inter-annotator agreement measured on two semantically and syntactically different groups of verbs – verbs of communication and exchange (85.9% and 78.5%, respectively). Based on the upper level semantic frames from the relation of ‘Inheritance’ built in FrameNet, the verbs of these eight groups have been classified into more coherent semantic classes. Moreover, frame elements from these upper level semantic frames have been assigned to valency complementations of the verbs of the listed groups as semantic roles. As in case of semantic frames, the achieved interannotator agreement concerning assigning frame elements measured on verbs of communication and exchange has been promising (95.6% and 91.2%, respectively). As a result, 1 270 lexical units pertaining to the verbs of communication, mental action, psych verbs, social interaction, verbs of exchange, motion, transport and location (2 129 Czech verbs in total if perfective and imperfective verbs being counted separately) have been classified into syntactically and semantically coherent classes and their valency complementations have been characterized by semantic roles adopted from the FrameNet lexical database.

Automatic acquisition of lexical knowledge is critical to a wide range of natural language processing tasks. Especially important is knowledge about verbs, which are the primary source of relational information in a sentence-the predicate-argument structure that relates an action or state to its participants (i.e., who did what to whom). In this work, we report on supervised learning experiments to automatically classify three major types of English verbs, based on their argument structure-specifically, the thematic roles they assign to participants. We use linguistically-motivated statistical indicators extracted from large annotated corpora to train the classifier, achieving 69.8% accuracy for a task whose baseline is 34%, and whose expert-based upper bound we calculate at 86.5%. A detailed analysis of the performance of the algorithm and of its errors confirms that the proposed features capture properties related to the argument structure of the verbs. Our results validate our hypotheses that knowledge about thematic relations is crucial for verb classification, and that it can be gleaned from a corpus by automatic means. We thus demonstrate an effective combination of deeper linguistic knowledge with the robustness and scalability of statistical techniques.

“…it is not the argument structure that triggers the intonational phrasing, but the [discourse, subclausal n.b.] relation of backgrounding.” (K. von Heusinger, 2007) Is it? If yes, how? The present paper, maintaining our attempts for applying Prague School's Topic-Focus Articulation (TFA) algorithm on the syntax-prosody interface of Romanian, proposes two comparative lines of investigation for the intonational focus assignment: (A) The TFA algorithm is improved at clause level with hints from Gussenhoven's SAAR (Sentence Accent Assignment Rule), then extended to inter-clause level, i.e. complex sentences. The new shape of the TFA algorithm is applied to compute the Topic-Focus values in the discursive context, while the information-structural (IS) spans of Theme(s)-Rheme(s) are detached, at clause level, as lowest-highest degrees of Communicative Dynamism (CD) vs. Systemic Order (SO). (B) The second approach we experiment for assigning intonational focus and phrasing is based on the combined and intensive use of discourse theories for computing the IS categories and structures: the Background-Kontrast entities (associated with the Prague School's Topic-Focus) are obtained with Asher's (1993) Segmented Discourse Representation Theory (SDRT) analysis, while Theme-Rheme structures within the finite clause are computed with Leong's (2004) Inference-Boundary (IB) algorithm (of Hallidayan inspiration), applied for the first time to Romanian. Furthermore, this second direction is inspired from and joins the IS-discourse theory proposed by Heusinger (2007), which relies on SDRT inter-clausal evolution of discourse variables for computing the Background-Kontrast. While maintaining the classical SDRT (including rhetorical) discourse relations at the inter-clause level, Heusinger introduces a set of IS-semantics relations, and hands down at sub-clause level the rhetorical and focus particle relations with significant role in intonational-prosodic phrasing. Examples of these two types of research are compared to a gold, intonationally annotated set of Romanian sentences, the proposed theoretical and procedural techniques aiming to balance the pessimistic-realistic view on prosody prediction that it is the speaker-presupposition (and hearer-accommodation) which determines the IS focal scopes, rather than the bare text.

Enriching learning materials with semantic roles

In this paper we present the first step in a larger series of experiments for the induction of predicate/argument structures.The structures that we are inducing are very similar to the conceptual structures that are used in Frame Semantics (such as FrameNet).Those structures are called messages and they were previously used in the context of a multi-document summarization system of evolving events.The series of experiments that we are proposing are essentially composed from two stages.In the first stage we are trying to extract a representative vocabulary of words.This vocabulary is later used in the second stage, during which we apply to it various clustering approaches in order to identify the clusters of predicates and arguments-or frames and semantic roles, to use the jargon of Frame Semantics.This paper presents in detail and evaluates the first stage.

Deaf users might find it difficult to navigate through websites with textual content which, for many of them, constitutes the written representation of a non-native oral language. With the aim of testing how the information structure could compensate for this difficulty, 27 prelingual deaf users of sign language were asked to search a set of headlines in a web newspaper where information structure and practice were manipulated. While practice did not affect deep structures (web content distributed through four layers of nodes), wide structures (web content concentrated in two layers) did facilitate users' performance in the last trial block and compromised it in the first trial block. It is argued that wide structures generate a textual information overload for deaf users, which decreases with practice. Thus, wide structures seem preferable for websites requiring frequent use, rather than for those intended for occasional interaction.

Human thought operates through associative processes in which concepts are activated through language. These concepts are organized around semantic frames (Fillmore, 1982), understood as conceptual structures that support the representation of specialized knowledge. Although frames play a key role in knowledge representation, systematic methods for their construction remain scarce and often overlook terminological and cultural dimensions. This study proposes constructing semantic frames differentiated by language and culture to compare how specialized concepts are lexicalized and thereby establish equivalences. The research focuses on the frame SUSTAINABLE FOREST MANAGEMENT in Spanish and French—an event of social and environmental significance whose terminology is underrepresented in current terminological resources. The frame-construction methodology, based on a corpus-driven approach (Biber, 2012), comprises six phases: (i) compilation of the FORESCOR corpus (2,124,653 words) from technical and legal forestry texts in Spanish and French; (ii) terminological extraction and lexical–semantic analysis of terms; (iii) identification and semantic annotation of argument structures to detect semantic relations (Sánchez-Cárdenas, 2024); (iv) generalization of semantic patterns and identification of phraseological equivalences; (v) construction of semantic frames in both languages; and (vi) comparison of the frames to identify interlinguistic equivalences, complemented by the application of distributional semantics techniques to FORESCOR (Cabezas-García, 2021). The analysis describes how lexicalized concepts are organized in each culture, establishes interlinguistic equivalences, and identifies cases of cultural anisomorphism. This work—transferable to other specialized domains—contributes to the development of effective terminographic resources for multilingual professional contexts. The results inform the design of a prototype Spanish–French forestry terminographic resource, intended to support the translation of forestry texts, within the framework of a broader ongoing study (FORESTERM: “Forestry Terminographic Resource”).

Frame Semantic Parsing (FSP) entails identifying predicates and labeling their arguments according to Frame Semantics. This paper investigates the use of In-Context Learning (ICL) with Large Language Models (LLMs) to perform FSP without model fine-tuning. We propose a method that automatically generates task-specific prompts for the Frame Identification (FI) and Frame Semantic Role Labeling (FSRL) subtasks, relying solely on the FrameNet database. These prompts, constructed from frame definitions and annotated examples, are used to guide six different LLMs. Experiments are conducted on a subset of frames related to violent events. The method achieves competitive results, with F<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</inf> scores of 94.3 for FI and 77.4 for FSRL. The findings suggest that ICL offers a practical and effective alternative to traditional fine-tuning for domain-specific FSP tasks.

This paper introduces a new, graph-based view of the data of the FrameNet project, which we hope will make it easier to understand the mixture of semantic and syntactic information contained in FrameNet annotation. We show how English FrameNet and other Frame Semantic resources can be represented as sets of interconnected graphs of frames, frame elements, semantic types, and annotated instances of them in text. We display examples of the new graphical representation based on the annotations, which combine Frame Semantics and Construction Grammar, thus capturing most of the syntax and semantics of each sentence. We consider how graph theory could help researchers to make better use of FrameNet data for tasks such as automatic Frame Semantic role labeling, paraphrasing, and translation. Finally, we describe the development of FrameNet-like lexical resources for other languages in the current Multilingual FrameNet project. which seeks to discover cross-lingual alignments, both in the lexicon (for frames and lexical units within frames) and across parallel or comparable texts. We conclude with an example showing graphically the semantic and syntactic similarities and differences between parallel sentences in English and Japanese. We will release software for displaying such graphs from the current data releases.

We present LAIR: A domain-specific language that enables users to specify actions to be taken upon meeting specific semantic frames in a text, in particular to rephrase and redact the textual content. While LAIR presupposes superficial knowledge of frames and frame semantics, it requires only limited prior programming experience. It neither contain scripting or I/O primitives, nor does it contain general loop constructions and is not Turing-complete. We have implemented a LAIR compiler and integrated it in a pipeline for automated redaction of web pages. We detail our experience with automated redaction of web pages for subjectively undesirable content; initial experiments suggest that using a small language based on semantic recognition of undesirable terms can be highly useful as a supplement to traditional methods of text sanitization.

The blocking problem caused by structural plane cutting is often the primary problem in evaluating the stability of rock mass engineering. The geometric parameters of the block structural plane are important research bases for analyzing block stability. In order to solve the problems of low efficiency, heavy workload, and high subjectivity in the traditional acquisition of structural plane occurrence information, taking the slope at the entrance of Suichang Gold Mine National Mine Park as the research background, this paper studies the rapid extraction method of structural plane information based on three-dimensional laser scanning data, and puts forward the process and data processing method of using a 3D laser scanner to collect slope block structural information to obtain point cloud data. Based on the coordinate projection principle and MATLAB, the rock slope block stability analysis CPG program was developed, which realized the whole process from data acquisition to block stability analysis. The research results are essential for identifying structural plane and block stability analysis and can provide theoretical and technical support for practical engineering applications.

Link prediction in Online Social Networks—OSNs—has been the focus of numerous studies in the machine learning community. A successful machine learning-based solution for this task needs to (i) leverage global and local properties of the graph structure surrounding links; (ii) leverage the content produced by OSN users; and (iii) allow their representations to change over time, as thousands of new links between users and new content like textual posts, comments, images and videos are created/uploaded every month. Current works have successfully leveraged the structural information but only a few have also taken into account the textual content and/or the dynamicity of network structure and node attributes. In this paper, we propose a methodology based on temporal graph neural networks to handle the challenges described above. To understand the impact of textual content on this task, we provide a novel pipeline to include textual information alongside the structural one with the usage of BERT language models, dense preprocessing layers, and an effective post-processing decoder. We conducted the evaluation on a novel dataset gathered from an emerging blockchain-based online social network, using a live-update setting that takes into account the evolving nature of data and models. The dataset serves as a useful testing ground for link prediction evaluation because it provides high-resolution temporal information on link creation and textual content, characteristics hard to find in current benchmark datasets. Our results show that temporal graph learning is a promising solution for dynamic link prediction with text. Indeed, combining textual features and dynamic Graph Neural Networks—GNNs—leads to the best performances over time. On average, the textual content can enhance the performance of a dynamic GNN by 3.1% and, as the collection of documents increases in size over time, help even models that do not consider the structural information of the network.

The Kazakh language is an under-researched Turkic language spoken in the Central Asian state of Kazakhstan and some neighbouring countries. While the grammar of this language is fairly well described, its information structural characteristics have not been examined in detail in the literature to date. This thesis aims to start filling this research gap by providing detailed descriptions of: the relation between information structure and word order; topic markers; and a pragmatically significant particle. Original, contextualised language examples are used to reject previous limited and rigid understanding of the relation between information structure and word order in Kazakh. It is shown that the information structural configurations of a Kazakh sentence are far more diverse than had been assumed. This conclusion is not only a revelation in its own right, but can also serve as a foundation for further research on the information structure of Kazakh, and other under-researched Turkic languages. This thesis also provides the first detailed descriptions of the three Kazakh topic markers. Numerous examples of their uses are presented in order to demonstrate the differences in these markers’ distribution and meaning. Several grammaticalisation processes related to these topic markers are revealed; it is proposed that these processes are currently at different stages of progress. Pragmatically significant particle goj is examined in detail for the first time: its distribution and meaning are illustrated with contextualised examples from various sources. It is posited that there are two syntactically diverse variants of this item which do, however, share the same existential semantics. The theoretical framework of Dynamic Syntax is employed throughout the thesis to underpin the first formal analyses of the phenomena under discussion.

The text analyses the relation between argument structure and information structure with “core dative verbs” in European Portuguese (EP), specifically in the constructions with heavy direct objects (DO). Developing some previous analysis , I will argue in favour of the proposal that EP has a basic construction of ditransitives where the DO is projected as complement of a low verbal category and the IO is projected in the specifier position of that verbal category. When the DP DO raises to a higher position because of Case reasons, the order V DO IO is obtained. With a heavy and focalized DO the order is V IO OD and that can be explained if the DP DO leaves a spelled-out copy. This approach makes discursive functional categories such that topic and focus in the left periphery of the Verbal Phrase unnecessary and reinforces the view that in languages like European Portuguese information structure relates, in a very strong way, Syntax and Phonology and Syntax and Prosody.