Abusive language detection in youtube comments leveraging replies as conversational context

Nowadays, social media experience an increase in hostility, which leads to many people suffering from online abusive behavior and harassment. We introduce a new publicly available annotated dataset for abusive language detection in short texts. The dataset includes comments from YouTube, along with contextual information: replies, video, video title, and the original description. The comments in the dataset are labeled as abusive or not and are classified by topic: politics, religion, and other. In particular, we discuss our refined annotation guidelines for such classification. We report a number of strong baselines on this dataset for the tasks of abusive language detection and topic classification, using a number of classifiers and text representations. We show that taking into account the conversational context, namely, replies, greatly improves the classification results as compared with using only linguistic features of the comments. We also study how the classification accuracy depends on the topic of the comment.

Personal Pronouns: Variation and Ambiguity

Personal pronouns are vague and highly versatile. In addition to their canonical functions as deictics and anaphors, they can be used to express meanings that go beyond morphosyntactic mapping and feature matching. Potential ambiguity is minimised by a variety of syntactic and extra-syntactic means, including the conversational context. Disambiguation through categorical morphological distinctions is rarely needed. Different non-canonical uses that may theoretically result in ambiguous utterances are presented to illustrate how speakers embrace variable pronoun choice that eludes prescriptive isomorphism, for the sake of expressivity and pragmatic meaning. An ‘Avoid Ambiguity’ principle is suggested for conversation that takes account of the benefits of linguistic variability, vagueness, and the situatedness of natural talk.

Literary device: allusion and its types

In this article, we explained to take a look at the meaning of allusion and how it can be used in both day to day conversation as well as how it can be used in literature. We enforced this understanding by taking a look at some examples of how allusion can be used in both of these contexts. In most cases, allusion is used to divert the mind to something which is not within the general context of the current conversation and is often left to the imagination of the listener or reader to create the reference for themselves. When used in writing, an allusion can make a piece appear less bland and much more artful creating a more interesting reading experience. Allusion can be used in both a conversational context or within written work as a literary device. Key words: allusion, literary, mythical, figure of speech, reader, historical

Conversational Disgust and Social Oppression

In recent years, philosophers have begun to uncover the role played by verbal conduct in generating oppressive social structures. I examine the oppressive illocutionary uses, and perlocutionary effects, of expressives: speech acts that are not truth-apt, merely expressing attitudes, such as desires, preferences, and emotions. Focusing on expressions of disgust in conversation, I argue for two claims: (1) that expressions of disgust can activate in the local, conversational context the oppressive power of the underlying structures of oppression; (2) that conversational expressions of disgust can, via the pragmatic process of presupposition accommodation, contribute to morally problematic cases of disgust contagion.

The effects of using two variants of one language on cognition: Evidence from bidialectalism and diglossia.

The question of whether and how bilingualism affects domain general cognition has been extensively debated. Less attention has been paid to the cognitive abilities of speakers of different variants of the same language, in linguistic situations such as bidialectalism and diglossia. Similarly to the bilingual situation, in bidialectalism and diglossia speakers need to use only one variant of the language in a given context. However, these situations provide fewer opportunities for mixing or switching between the variants, potentially leading to different domain general cognitive outcomes than those reported in bilingualism. Here we review the available evidence on the effects of bidialectalism and diglossia on cognition, and evaluate it in relation to theories of the effects of bilingualism on cognition. We conclude that investigations of bilingualism, bidialectalism and diglossia must take into account the conversational context and, in particular, the opportunities for language switching that this affords.

Form and Function of Aizuchi Japenese Native Speaker in Inaka Ni Tomarou! TV Serial

This study is intended to describe the form of aizuchi in the TV series Inaka Ni Tomarou! and find out the function of Aizuchi's speech based on the conversation context. The research data are the forms of aizuchi, which are used in a conversational context. This type of research is a qualitative descriptive study. The researchers found six forms of aizuchi: short speech, interjection speech, interjection, and short utterances, repeated short utterances, repeated speech partner utterance, short utterances, and repetition of speech partners. In the function, Aizuchi has seven functions: a continuer signal, an understanding signal, an approval signal, a signal indicating emotion, a signal to confirm, a rejection signal, and a filling signal.

WHOSE SHAREDPLANS? SCRIPTS, COLLABORATION, AND FEMINIST AI RESEARCH

This paper examines a network of women in AI research who together expanded the range of methodologies and disciplines usually included in AI in the 1980s and 1990s. In particular, Barbara Grosz and Candace Sidner’s concept of $2 offered a way to model conversational context and collaboration in multi-agent AI environments. Drawing on archival work, interviews, conference proceedings, white papers, and departmental reports, I consider the cultural, institutional, and intellectual forces that shaped this network and their research. Using a technofeminist framework (Wajcman 2004; Haraway 1990) and borrowing from Michelle Murphy’s (2012) concept of protocol feminism, this paper examines their “feminist AI protocol.” I outline on one hand an assemblage of techniques, values, methods, and practices that illustrate a protocol rooted in community, interdisciplinarity, and care; these researchers formalized human-computer dialogue as fundamentally collaborative, grounding their approach in the diverse goals and desires of real users. I argue their philosophy of “language as action” mirrors ideas circulating in feminist and critical STS simultaneously. On the other hand, this network of researchers did so from within a particular set of cultural and epistemological parameters of their computer science departments. The research practices of this network offer an opportunity to consider the limits of any feminist AI protocol without a deeper commitment to feminist epistemologies. There remains an urgent need to reflect on how to build feminist AI technologies that make room for and include many different standpoints.

Thinking Globally, Acting Locally: Distantly Supervised Global-to-Local Knowledge Selection for Background Based Conversation

Background Based Conversation (BBCs) have been introduced to help conversational systems avoid generating overly generic responses. In a BBC, the conversation is grounded in a knowledge source. A key challenge in BBCs is Knowledge Selection (KS): given a conversational context, try to find the appropriate background knowledge (a text fragment containing related facts or comments, etc.) based on which to generate the next response. Previous work addresses KS by employing attention and/or pointer mechanisms. These mechanisms use a local perspective, i.e., they select a token at a time based solely on the current decoding state. We argue for the adoption of a global perspective, i.e., pre-selecting some text fragments from the background knowledge that could help determine the topic of the next response. We enhance KS in BBCs by introducing a Global-to-Local Knowledge Selection (GLKS) mechanism. Given a conversational context and background knowledge, we first learn a topic transition vector to encode the most likely text fragments to be used in the next response, which is then used to guide the local KS at each decoding timestamp. In order to effectively learn the topic transition vector, we propose a distantly supervised learning schema. Experimental results show that the GLKS model significantly outperforms state-of-the-art methods in terms of both automatic and human evaluation. More importantly, GLKS achieves this without requiring any extra annotations, which demonstrates its high degree of scalability.

Modality-Balanced Models for Visual Dialogue

The Visual Dialog task requires a model to exploit both image and conversational context information to generate the next response to the dialogue. However, via manual analysis, we find that a large number of conversational questions can be answered by only looking at the image without any access to the context history, while others still need the conversation context to predict the correct answers. We demonstrate that due to this reason, previous joint-modality (history and image) models over-rely on and are more prone to memorizing the dialogue history (e.g., by extracting certain keywords or patterns in the context information), whereas image-only models are more generalizable (because they cannot memorize or extract keywords from history) and perform substantially better at the primary normalized discounted cumulative gain (NDCG) task metric which allows multiple correct answers. Hence, this observation encourages us to explicitly maintain two models, i.e., an image-only model and an image-history joint model, and combine their complementary abilities for a more balanced multimodal model. We present multiple methods for this integration of the two models, via ensemble and consensus dropout fusion with shared parameters. Empirically, our models achieve strong results on the Visual Dialog challenge 2019 (rank 3 on NDCG and high balance across metrics), and substantially outperform the winner of the Visual Dialog challenge 2018 on most metrics.