Anaphora Resolution Research Articles

Can Anaphora Resolution Improve Extractive Query-Focused Multi-Document Summarization?

Can Anaphora Resolution Improve Extractive Query-Focused Multi-Document Summarization?

Winograd schemata and other datasets for anaphora resolution in Hungarian

AbstractThe Winograd Schema Challenge (WSC, proposed by Levesque, Davis & Morgenstern 2012) is considered to be the novel Turing Test to examine machine intelligence. Winograd schema questions require the resolution of anaphora with the help of world knowledge and commonsense reasoning. Anaphora resolution is itself an important and difficult issue in natural language processing, therefore, many other datasets have been created to address this issue. In this paper we look into the Winograd schemata and other Winograd-like datasets and the translations of the schemata to other languages, such as Chinese, French and Portuguese. We present the Hungarian translation of the original Winograd schemata and a parallel corpus of all the translations of the schemata currently available. We also adapted some other anaphora resolution datasets to Hungarian. We aim to discuss the challenges we faced during the translation/adaption process.

Референциальные особенности текстов нарративного жанра при освоении первого и второго языка

The paper regards the process of the development of several features of the narrative speech genre in the discourse field of Russian culture while learning Russian as a second language. Referentiality in general linguistic sense is an area concerning naming of objects and characters by language means. In this paper we analyse the problems of the referential choice (what factors influence the speaker’s choice of a referential expression) and anaphora resolution (how the listener/reader binds an anaphoric pronoun to a coreferential noun phrase). The main goal of our paper is to study referential features of narratives produced by foreign students acquiring Russian as a second language (RSL), compared with adult Russian speakers and children acquiring Russian as a first language (RFL). Also, we describe referential mistakes made by RSL learners. The analysis of the data demonstrates that the main referential features (abilities to make the right referential choices, form anaphoric chains, etc.) are related to the age-specific cognitive development and do not depend on the native language. Both groups of adults show similar patterns in using nouns and pronouns for the introduction or reintroduction of a referent and for the topic maintenance, and build continuous anaphoric chains. Particular anaphoric mistakes of RSL learners are related to the level of their grammar skills and differ from the mistakes made by RFL children: while children’s mistakes base on the cognitive factor, RSL students’ mistakes are mostly grammatical.

Computational Models of Anaphora

Interpreting anaphoric references is a fundamental aspect of our language competence that has long attracted the attention of computational linguists. The appearance of ever-larger anaphorically annotated data sets covering more and more anaphoric phenomena in ever-greater detail has spurred the development of increasingly more sophisticated computational models; as a result, the most recent state-of-the-art neural models are able to achieve impressive performance by leveraging linguistic, lexical, discourse, and encyclopedic information. This article provides a thorough survey of anaphora resolution (coreference) throughout this development, reviewing the available data sets and covering both the preneural history of the field and—in more detail—current neural models, including research on less-studied aspects of anaphoric interpretation such as bridging reference resolution and discourse deixis interpretation.

Researches Advanced in the Development and Application of Information Extraction

Natural language processing (NLP) is an interdisciplinary subject of linguistics and computer science, which processes the language used for human communication into a machine language that can be understood by machines.The task of information extraction is to obtain target information accurately and quickly from a large amount of data and improve the utilization of information. With the rapid development of Internet applications, how to quickly and accurately analyze the really useful information in these text data is particularly critical and urgent. Information extraction has become an important branch of natural language processing. Thanks to the rapid development of deep learning, the performance of information extraction has made breakthroughs in recent years. In this paper, based on the detailed literature analysis, this paper first review the development of information extraction. Secondly, the research progress of key technologies of information extraction is summarized from four aspects: named entity recognition, anaphora resolution, relationship extraction and event extraction. Finally, we analyze some main problems of information extraction and predict the research trend of information extraction.

Direct object anaphora resolution in L1 English-L2 Spanish

Abstract This study investigates the written production of direct object (DO) anaphora in L2 Spanish by adult English native speakers. It focuses on the use of clitics and full DPs in topic continuity, where highly salient topics are typically recovered by clitics in native Spanish. A combination of pragmatic factors in subject anaphora resolution (referential ambiguity and distance) and morphosyntactic factors in the acquisition of clitics (gender and animacy) was analyzed in 5 subcorpora from the CEDEL2 (Corpus Escrito del Español L2). Following Learner Corpus Research methodology, 773 anaphoric DOs and their antecedents were tagged across 150 texts (from low-intermediate to upper-advanced proficiency). Results showed that as learner proficiency increases, the predominant anaphoric strategy shifts from redundant DP overuse to felicitous pronominalization. Clitic avoidance is the general strategy intimately related to clitics’ morphosyntactic deficits at advanced levels, but also to pragmatic principles supporting the Pragmatic Principles Violation Hypothesis (Lozano, 2016).

Specific pattern of linguistic impairment in Parkinson's disease patients with subjective cognitive decline and mild cognitive impairment predicts dementia.

Parkinson's disease patients with subjective cognitive decline (PD-SCD) and mild cognitive impairment (PD-MCI) have an increased risk of dementia (PDD). Thus, the identification of early cognitive changes that can be useful predictors of PDD is a highly relevant challenge. Posterior cortically based functions, including linguistic processes, have been associated with PDD. However, investigations that have focused on linguistic functions in PD-MCI are scarce and none of them include PD-SCD patients. Our aim was to study language performance in PD-SCD and PD-MCI. Moreover, language subcomponents were considered as predictors of PDD. Forty-six PD patients and twenty controls were evaluated with a neuropsychological protocol. Patients were classified as PD-SCD and PD-MCI. Language production and comprehension was assessed. Follow-up assessment was conducted to a mean of 7.5 years after the baseline. PD-MCI patients showed a poor performance in naming (actions and nouns), action generation, anaphora resolution and sentence comprehension (with and without center-embedded relative clause). PD-SCD showed a poor performance in action naming and action generation. Deficit in action naming was an independent risk factor for PDD during the follow-up. Moreover, the combination of deficit in action words and sentence comprehension without a center-embedded relative clause was associated with a greater risk. The results are of relevance because they suggest that a specific pattern of linguistic dysfunctions, that can be present even in the early stages of the disease, can predict future dementia, reinforcing the importance of advancing in the knowledge of linguistic dysfunctions in predementia stages of PD.

The processing of the ambiguous object pronoun in German monolingual and highly proficient L2 speakers with L1 Georgian

Aims and objectives: The current study examines the role of grammatical and positional parallelism on ambiguous German pronoun resolution in monolingual speakers and highly proficient L2 speakers with L1 Georgian. In particular, the study asks whether an object pronoun ihn in a sentence initial position could refer to the non-subject antecedent in a preceding subject–verb–object sentence. Furthermore, it investigates whether L1 and L2 speakers show similar preferences when resolving ambiguous object pronouns. Methodology: Two visual world eye-tracking and two offline experiments were conducted with L1 and L2 speakers of German. During the eye-tracking session, the task was simply to look and listen, while in the offline test, participants were required to decide on the assumed referent of the ambiguous pronoun. Data and analysis: Linear-mixed effect models were applied to the eye-movement data. Two-sample t tests were used in the analysis of the offline data. Findings and conclusions: Eye-movement results revealed a bias toward the subject antecedent in L1 speakers, while the L2 speakers showed a non-subject preference when resolving the referent of the object pronoun. The offline experiments supported the eye-tracking findings of both groups, particularly as regards the role of grammatical parallelism in the resolution of ambiguous pronouns by the L2 speakers. These findings are interpreted in the context of language-specific processing strategies in anaphora resolution. It appears that L1 and L2 speakers may be using different information-structural cues to resolve ambiguous pronouns. Concretely, a less strongly internalized knowledge of the target language or the influence of the first language may produce alternative processing mechanisms in L2 speakers. Originality: This study fills the research gap on ambiguous object pronoun resolution of L1 and L2 speakers of German. Implications: The findings of this study add additional information to the growing body of research into pronoun resolution. It focuses on the resolution of the German object pronoun within an information-structural framework of language processing in L1 and L2.

Second chances in antecedent retrieval: The processing of reflexives in two types of reconstruction environments

AbstractA body of psycholinguistic research shows that structural constraints play a large role in guiding anaphora resolution in the early moments of processing (Nicol and Swinney 1989; Harris et al. 2000; Sturt 2003; Kazanina et al. 2007; Xiang et al. 2009; Chow et al. 2014). Omaki (2010) and Omaki et al. (2019) report on an interesting case where reflexives in wh-predicate fronting constructions launch a search that is not structurally guided. We further investigate this phenomenon, by asking whether non-structurally guided retrievals of this sort result in comprehenders ever committing to ungrammatical antecedents, perhaps as a ‘lingering’ interpretation of the sort found with garden path sentences (Christianson et al. 2001; Ferreira et al. 2001; Slattery et al. 2013). In two forced-choice studies, we find evidence that ungrammatical dependencies resulting from a non-structural search are sometimes maintained in offline comprehension, particularly with a more demanding task. We then probe the incremental processing that follows non-structurally guided retrieval, asking if and when the processor initiates a renewed search. In a self-paced reading experiment, we show that the processor continues its search for an antecedent very soon after retrieving a non-structurally guided antecedent. Surprisingly, however, we found a similar processing profile in cases where a structurally licensed antecedent was already encountered. While it has been recently shown that cataphoric pronouns persist in an antecedent search after a failed retrieval (Giskes and Kush 2021), our results suggest that when reflexives locate a preceding antecedent – by either a structural or non-structural search – this does not terminate further consideration of a different dependency. We consider these data in light of the comparison between cataphoric elements and wh-fillers in launching an active search to complete a dependency.

Input Utterance Complementation Method by Anaphora Resolution for Spontaneous Utterances on Spoken Dialog Systems

Input Utterance Complementation Method by Anaphora Resolution for Spontaneous Utterances on Spoken Dialog Systems

Anaphora Resolution in Thai EDU Segmentation

Anaphora Resolution in Thai EDU Segmentation

A Review on Methods to Solve Polysemy Problem in WSD Occurring while Processing Marathi Text

Word Sense Disambiguation (WSD) is an open problem in computational linguistics concerned with identifying which sense of a word is used in a sentence. It basically resolves the ambiguity that arises when determining the meaning of the same word used in a particular context in different situations. The solution to this issue impacts other computer-related writing, such as discourse, improving relevance of search engines, anaphora resolution, coherence and inference. Marathi is an official language of Maharashtra, India, and co-official Language of Goa & Union territory of Dadra & Nagar Haveli and Daman & Diu with more than an 83 million Native speaker whereas 12 million speaks Marathi as their second language. However, the language barrier is impeding the information technology revolution’s progress. Thus, there is a requisite for competent measures to carry out natural language processing (NCP) through computer processing so that computers can interact through language like Marathi and handle by users who know regional Language. This paper provides a review of the methods for WSD and present modified evaluation. The methods, algorithms, and techniques of Word Sense Disambiguation are examined in this work. We aim to elaborate on resources that will be useful for working in Marathi Word Sense Disambiguation in this paper.

Experiences on the Improvement of Logic-Based Anaphora Resolution in English Texts

Anaphora resolution is a crucial task for information extraction. Syntax-based approaches are based on the syntactic structure of sentences. Knowledge-poor approaches aim at avoiding the need for further external resources or knowledge to carry out their task. This paper proposes a knowledge-poor, syntax-based approach to anaphora resolution in English texts. Our approach improves the traditional algorithm that is considered the standard baseline for comparison in the literature. Its most relevant contributions are in its ability to handle differently different kinds of anaphoras, and to disambiguate alternate associations using gender recognition of proper nouns. The former is obtained by refining the rules in the baseline algorithm, while the latter is obtained using a machine learning approach. Experimental results on a standard benchmark dataset used in the literature show that our approach can significantly improve the performance over the standard baseline algorithm used in the literature, and compares well also to the state-of-the-art algorithm that thoroughly exploits external knowledge. It is also efficient. Thus, we propose to use our algorithm as the new baseline in the literature.

Pronominal anaphora resolution in Polish: Investigating online sentence interpretation using eye-tracking.

The mechanism of anaphora resolution is subject to large cross-linguistic differences. The most likely reason for this is the different sensitivity of pronouns to the range of factors that determine their reference. In the current study, we explored the mechanism of anaphora resolution in Polish. First, we explored preferences in the interpretation of null and overt pronouns in ambiguous sentences. More specifically, we investigated whether Polish speakers prefer to relate overt pronouns to subject or object antecedents. Subsequently, we tested the consequences of violating this bias when tracing the online sentence-interpretation process using eye-tracking. Our results show that Polish speakers have a strong preference for interpreting null pronouns as referring to subject antecedents and interpreting overt pronouns as referring to object antecedents. However, in online sentence interpretation, only overt pronouns showed sensitivity to a violation of the speaker's preference for a pronoun-antecedent match. This suggests that null pronoun resolution is more flexible than overt pronoun resolution. Our results indicate that it is much easier for Polish speakers to shift the reference of a null pronoun than an overt one whenever a pronoun is forced to refer to a less-preferred antecedent. These results are supported by naturalness ratings, which showed that null pronouns are considered equally natural regardless of their reference, while overt pronouns referring to subject antecedents are rated as considerably less natural than those referring to object antecedents. To explain this effect, we propose that the interpretation of null and overt pronouns is sensitive to different factors which determine their reference.

Observers use gesture to disambiguate contrastive expressions of preference

AbstractWe present two studies exploring how participants respond when a speaker contrasts two ideas, then expresses an ambiguous preference towards one of them. Study 1 showed that, when reading a speaker’s preference as text, participants tended to choose whatever was said last as matching the speaker’s preference, reflecting the recent-mention bias of anaphora resolution. In Study 2, we asked whether this pattern changed for audio versions of our stimuli. We found that it did not. We then asked whether observers used gesture to disambiguate the speaker’s preference. Participants watched videos in which two statements were spoken. Co-speech gestures were produced during each statement, in two different locations. Next, an ambiguous preference for one option was spoken. In ‘gesture disambiguating’ trials, this statement was accompanied by a gesture in the same spatial location as the gesture accompanying the first statement. In ‘gesture non-disambiguating’ trials, no third gesture occurred. Participants chose the first statement as matching the speaker’s preference more often for gesture disambiguating compared to non-disambiguating trials. Our findings add to the literature on resolution of ambiguous anaphoric reference involving concrete entities and discourse deixis, and we extend this literature to show that gestures indexing abstract ideas are also used during discourse comprehension.

Pattern-Based Syntactic Simplification of Compound and Complex Sentences

With the advent of new technologies, simplifying the text automatically has been very popular and of high importance among natural language researchers during the last decade. The predominant research done in the area of Automatic Sentence Simplification(ASS) is inclined to either lexical or syntactical simplification of sentences. From the literature survey, it is observed that existing research in lexical simplification makes use of word substitution technique. This causes word sense ambiguity in cases where the word synonyms are not appropriate for a sentence in the given context. In contrast syntactical simplification though is accurate and applicable to Natural Language Processing (NLP) tasks requires tremendous efforts to construct rules for a given domain. The research proposes a framework called, Pattern-based Automatic Syntactic Simplification(PASS) which identifies sentences and applies rules based on grammatical patterns to simplify the sentences thereby making it more generic for NLP tasks. PASS is evaluated by human experts to rate the usefulness of the framework based on fluency, adequacy and simplicity of the sentences. Furthermore, the framework is automatically evaluated with the available online corpus using automatic metrics of SARI, BLEU, and FKGL. The proposed approach generates promising results in the field of ASS and could be used as a preliminary module for NLP tasks as well as other natural language-related applications like summarization, anaphora resolution, question-answering, and many more.

Anaphora resolution in L2 and 2L1 Hungarian

The paper tests the antecedent preferences for the null pronoun, the personal pronoun and the demonstrative subject in L2 and 2L1 Hungarian (with Romanian as the other language), by means of an offline picture selection task. In the null subject condition, both groups show subject bias, confirming previous hypotheses and research. An evident object bias is observable with the demonstrative, again confirming the low accessibility status of demonstratives. The results differ in the case of the overt personal pronoun subject. While it can have both topic shift and topic continuity interpretation in both languages, previous research suggests that in Romanian the former is favoured over the latter, whereas the reverse is true for Hungarian. Whereas the L2 group shows no bias in the personal pronoun condition, thus possibly showing transfer from L1, the 2L1 group shows a strong subject bias, placing the personal pronoun on equal footing with the null pronoun.

A resolução de ambiguidade anafórica em inglês, espanhol e português

ABSTRACT This study aims to determine how (a) ambiguity resolution differs in English, Spanish, and Portuguese; (b) L2 learners are influenced by their L1 regarding ambiguity resolution; and (c) the order of the clauses, (d) the salience of the anaphor, or (e) the choice of verbs may affect this process. A total of 181 people answered an online survey and selected the antecedent of the anaphor for 16 ambiguous sentences. The Position of Antecedent Strategy was tested for the three languages and the conclusion was that it applies to European and Brazilian Portuguese, but not for Spanish or English. While Spanish and English native speakers tend to interpret the subject of the subordinate clause as corresponding to the subject of the main clause, learners are influenced by their L1 and seem uncertain especially on how to resolve pronominal cataphora. Finally, we have concluded that semantics has a great influence on the resolution of ambiguous anaphora in these languages, and we suggest that it should be taken in consideration during the design of a study like this one.

Pseudo Zero Pronoun Resolution Improves Zero Anaphora Resolution

Masked language models (MLMs) have contributed to drastic performance improvements with regard to zero anaphora resolution (ZAR). To further improve this approach, in this study, we made two proposals. The first is a new pretraining task that trains MLMs on anaphoric relations with explicit supervision, and the second proposal is a new finetuning method that remedies a notorious issue, the pretrain-finetune discrepancy. Our experiments on Japanese ZAR demonstrated that our two proposals boost the state-of-the-art performance, and our detailed analysis provides new insights on the remaining challenges.

Anaphora Resolution in English Ambiguous Sentences by ESL Learners

Anaphora Resolution in English Ambiguous Sentences by ESL Learners