| { |
| "paper_id": "2020", |
| "header": { |
| "generated_with": "S2ORC 1.0.0", |
| "date_generated": "2023-01-19T15:40:29.856403Z" |
| }, |
| "title": "Knowledge Graphs meet Moral Values", |
| "authors": [ |
| { |
| "first": "Ioana", |
| "middle": [], |
| "last": "Hulpus", |
| "suffix": "", |
| "affiliation": { |
| "laboratory": "", |
| "institution": "University of Mannheim", |
| "location": {} |
| }, |
| "email": "" |
| }, |
| { |
| "first": "Jonathan", |
| "middle": [], |
| "last": "Kobbe", |
| "suffix": "", |
| "affiliation": { |
| "laboratory": "", |
| "institution": "University of Mannheim", |
| "location": {} |
| }, |
| "email": "" |
| }, |
| { |
| "first": "Heiner", |
| "middle": [], |
| "last": "Stuckenschmidt", |
| "suffix": "", |
| "affiliation": { |
| "laboratory": "", |
| "institution": "University of Mannheim", |
| "location": {} |
| }, |
| "email": "" |
| }, |
| { |
| "first": "Graeme", |
| "middle": [], |
| "last": "Hirst", |
| "suffix": "", |
| "affiliation": { |
| "laboratory": "", |
| "institution": "University of Toronto", |
| "location": {} |
| }, |
| "email": "" |
| } |
| ], |
| "year": "", |
| "venue": null, |
| "identifiers": {}, |
| "abstract": "Operationalizing morality is crucial for understanding multiple aspects of society that have moral values at their core-such as riots, mobilizing movements, public debates, etc. Moral Foundations Theory (MFT) has become one of the most adopted theories of morality partly due to its accompanying lexicon, the Moral Foundation Dictionary (MFD), which offers a base for computationally dealing with morality. In this work, we exploit the MFD in a novel direction by investigating how well moral values are captured by KGs. We explore three widely used KGs, and provide conceptlevel analogues for the MFD. Furthermore, we propose several Personalized PageRank variations in order to score all the concepts and entities in the KGs with respect to their relevance to the different moral values. Our promising results help to progress the operationalization of morality in both NLP and KG communities.", |
| "pdf_parse": { |
| "paper_id": "2020", |
| "_pdf_hash": "", |
| "abstract": [ |
| { |
| "text": "Operationalizing morality is crucial for understanding multiple aspects of society that have moral values at their core-such as riots, mobilizing movements, public debates, etc. Moral Foundations Theory (MFT) has become one of the most adopted theories of morality partly due to its accompanying lexicon, the Moral Foundation Dictionary (MFD), which offers a base for computationally dealing with morality. In this work, we exploit the MFD in a novel direction by investigating how well moral values are captured by KGs. We explore three widely used KGs, and provide conceptlevel analogues for the MFD. Furthermore, we propose several Personalized PageRank variations in order to score all the concepts and entities in the KGs with respect to their relevance to the different moral values. Our promising results help to progress the operationalization of morality in both NLP and KG communities.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Abstract", |
| "sec_num": null |
| } |
| ], |
| "body_text": [ |
| { |
| "text": "Many of the choices that we make in daily life, such as political stance or position in debates on ideological topics, are influenced by our moral values (Sagi and Dehghani, 2014; Wolsko et al., 2016; Amin et al., 2017) . Besides, moral values and moral judgments are central to decision making and cultural cohesion . The last years have seen an increasing interest in operationalizing the concept of morality as defined by psychologists, particularly from the NLP, social media, and communication communities, into an effort of extracting the latent moral dimension of texts.", |
| "cite_spans": [ |
| { |
| "start": 154, |
| "end": 179, |
| "text": "(Sagi and Dehghani, 2014;", |
| "ref_id": "BIBREF21" |
| }, |
| { |
| "start": 180, |
| "end": 200, |
| "text": "Wolsko et al., 2016;", |
| "ref_id": "BIBREF23" |
| }, |
| { |
| "start": 201, |
| "end": 219, |
| "text": "Amin et al., 2017)", |
| "ref_id": "BIBREF0" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "Tweets Araque et al., 2020) , newspaper articles (Bowman et al., 2014) , as well as scientific articles (Clifford and Jerit, 2013) or This work is licensed under a Creative Commons Attribution 4.0 International License. License details: http: //creativecommons.org/licenses/by/4.0/. religious sermons (Graham et al., 2009) have been targeted for moral analysis. Lately, this line of research has been widely motivated by the rise of social media campaigns such as #BlackLivesMatter and #MeToo, which have a very strong moral load. People take different stances with respect to such matters depending on their understanding and hierarchy of moral values, and can lead to clashes of visions even within the same culture. The general assumption is that the words used in discussions on such topics reveal the moral values of the discussants.", |
| "cite_spans": [ |
| { |
| "start": 7, |
| "end": 27, |
| "text": "Araque et al., 2020)", |
| "ref_id": "BIBREF1" |
| }, |
| { |
| "start": 49, |
| "end": 70, |
| "text": "(Bowman et al., 2014)", |
| "ref_id": "BIBREF3" |
| }, |
| { |
| "start": 104, |
| "end": 133, |
| "text": "(Clifford and Jerit, 2013) or", |
| "ref_id": "BIBREF5" |
| }, |
| { |
| "start": 301, |
| "end": 322, |
| "text": "(Graham et al., 2009)", |
| "ref_id": "BIBREF10" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "One of the most widely adopted theories of morality is the Moral Foundations Theory (MFT) (Haidt and Joseph, 2004; Graham et al., 2013) . MFT proposes at least five moral foundations, each one consisting on the one side of virtues and on the other side of vices: (1) the care/harm foundation which deals with the sensitivity towards the suffering of others; (2) the fairness/cheating foundation covering aspects of reciprocity and motivations to be fair; (3) the loyalty / betrayal foundation covering aspects of in-group cooperation, and the intuition of being loyal to one's group; (4) the authority/subversion foundation which is related to the innate intuition of endorsing hierarchies that we find just; (5) the purity / degradation foundation which deals with our innate drive of preferring cleanliness of body and soul over hedonism.", |
| "cite_spans": [ |
| { |
| "start": 90, |
| "end": 114, |
| "text": "(Haidt and Joseph, 2004;", |
| "ref_id": "BIBREF11" |
| }, |
| { |
| "start": 115, |
| "end": 135, |
| "text": "Graham et al., 2013)", |
| "ref_id": "BIBREF9" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "The Moral Foundations Dictionary (MFD) (Graham et al., 2009) has been proposed as a lexicon to guide the assessment of the moral foundations of the MFT in texts. It consists of a set of words and lemmas for each vice and virtue of each foundation and has become an essential resource for operationalizing moral values.", |
| "cite_spans": [ |
| { |
| "start": 39, |
| "end": 60, |
| "text": "(Graham et al., 2009)", |
| "ref_id": "BIBREF10" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "Nevertheless, being a word-level lexicon, this resource comes with several limitations. First, the natural ambiguity of language means that some of the words and lemmas provided can have other meanings than the ones related to the moral foundations. For instance, the stem subver* covers the ambiguous word subversion, which besides the meaning related to the authority / subversion foundation, it is used with a completely different meaning in software development.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "Second, it contains a limited set of lemmas and words, particularly focusing on those whose main meaning is the one related to the corresponding moral foundation. This, on the one side, means that more ambiguous synonyms are not contained, and on the other side, it means that very specific rarely used words are contained. These effects lead to a high precision with poor recall, which is not necessarily the preferred strategy in many scenarios.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "Moreover, it only contains uni-grams, and the entries are associated with either vice or virtue, without a score for the strength of the association. However, this lexicon has been widely exploited lately, and several approaches have been proposed to overcome these weaknesses, for instance, by extending it or by its projection into continuous spaces.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "In this work, we investigate a new direction, that of projecting the MFD lexicon on knowledge graphs (KGs) with the purpose of scoring all entities and concepts therein with respect to their relevance for each moral foundation. We envision multiple benefits from this endeavor. First, it overcomes the ambiguity and incompleteness limitations. Second, entities and concepts in KGs oftentimes strongly relate to moral values. For example, Rebecca Reichmann Tavares is a UN diplomat promoting race relations and human rights, a position highly related to the Fairness/Cheating moral foundation. Similarly, the concept History of Human Rights is also highly related to the same moral foundation. Such concepts and entities are not part of the MFD, but their mentioning can help the detection of moral foundations expressed by the texts.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "Third, the usefulness of KGs for many tasks resides in the fact that they provide factual knowledge such as relations between entities. Still, an important drawback of current KGs is their weak representation of common sense knowledge. For example, the fact that the concept of \"crime\" is generally bad and undesirable cannot be derived from current KGs although such common sense knowledge is crucial, for example, for understand-ing arguments: To understand that an argument claiming that racism leads to crime is an argument against racism, it is necessary to understand that crime is bad or a vice. In this work, we propose a means of adding a moral dimension to KGs, and hence extend them with common-sense \"intuition\" of morality 1 .", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "The main target of this work is to investigate how the moral foundations characterized by the MFD are captured in KGs, with the purpose of scoring each entity and concept in the KGs with respect to their relevance to the moral foundations. Therefore, we are particularly interested in the way previous literature uses the MFD. Some works (Graham et al., 2009; Clifford and Jerit, 2013; Teernstra et al., 2016; Lin et al., 2018) simply use MFD counts either on their own or as features in supervised classifiers for determining the moral values expressed by text. Works that try to overcome the issues related to the simple counts of lexicon hits embed the moral values in continuous spaces. and Kaur and Sasahara (2016) generate vectors for words based on a Latent Semantic Analysis (LSA) (Deerwester et al., 1990 ) methodology. Then, for each moral value, a vector in the same space is obtained by adding up all the vectors of the modeled lexicon words. , Nokhiz and Li (2017) and Xie et al. (2019) use Word2Vec (Mikolov et al., 2013) or GloVe (Pennington et al., 2014) models to embed the words of the lexicon, and then either aggregate these vectors to produce a vector representation of each moral value, or they keep each of these vectors separate and use different strategies for determining the moral values expressed by novel embedded texts (such as k-nearest neighbor).", |
| "cite_spans": [ |
| { |
| "start": 338, |
| "end": 359, |
| "text": "(Graham et al., 2009;", |
| "ref_id": "BIBREF10" |
| }, |
| { |
| "start": 360, |
| "end": 385, |
| "text": "Clifford and Jerit, 2013;", |
| "ref_id": "BIBREF5" |
| }, |
| { |
| "start": 386, |
| "end": 409, |
| "text": "Teernstra et al., 2016;", |
| "ref_id": "BIBREF22" |
| }, |
| { |
| "start": 410, |
| "end": 427, |
| "text": "Lin et al., 2018)", |
| "ref_id": "BIBREF15" |
| }, |
| { |
| "start": 695, |
| "end": 719, |
| "text": "Kaur and Sasahara (2016)", |
| "ref_id": "BIBREF14" |
| }, |
| { |
| "start": 789, |
| "end": 813, |
| "text": "(Deerwester et al., 1990", |
| "ref_id": "BIBREF6" |
| }, |
| { |
| "start": 957, |
| "end": 977, |
| "text": "Nokhiz and Li (2017)", |
| "ref_id": "BIBREF18" |
| }, |
| { |
| "start": 1013, |
| "end": 1035, |
| "text": "(Mikolov et al., 2013)", |
| "ref_id": "BIBREF17" |
| }, |
| { |
| "start": 1045, |
| "end": 1070, |
| "text": "(Pennington et al., 2014)", |
| "ref_id": "BIBREF19" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Related Work", |
| "sec_num": "2" |
| }, |
| { |
| "text": "Other works that go into a direction more related to ours extend the MFD lexicon by using Word-Net (Rezapour et al., 2019; Araque et al., 2020) , and then manually curate the results with human annotators (Rezapour et al., 2019) or extend the annotations with values for valence and arousal (Araque et al., 2020) . However, the methods differ from our approach as the results of these works are still word-level lexicons, while our moral value relevance annotation is done at a synset rather than word level. For example, MoralStrength (Araque et al., 2020) is built by manually expanding the MFD with new words from the same WordNet synsets.", |
| "cite_spans": [ |
| { |
| "start": 99, |
| "end": 122, |
| "text": "(Rezapour et al., 2019;", |
| "ref_id": "BIBREF20" |
| }, |
| { |
| "start": 123, |
| "end": 143, |
| "text": "Araque et al., 2020)", |
| "ref_id": "BIBREF1" |
| }, |
| { |
| "start": 205, |
| "end": 228, |
| "text": "(Rezapour et al., 2019)", |
| "ref_id": "BIBREF20" |
| }, |
| { |
| "start": 291, |
| "end": 312, |
| "text": "(Araque et al., 2020)", |
| "ref_id": "BIBREF1" |
| }, |
| { |
| "start": 536, |
| "end": 557, |
| "text": "(Araque et al., 2020)", |
| "ref_id": "BIBREF1" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Related Work", |
| "sec_num": "2" |
| }, |
| { |
| "text": "The only work that acknowledges that entities from Wikipedia/DBpedia also carry a moral load is that of Lin et al. (2018) . They use Wikipedia abstracts and DBpedia properties in order to generate features of the entities mentioned in text and use these features together with the textual features in order to classify the texts based on their moral values. However, they do not provide a moral value score of the specified entities, and they do not look into linking the MFD to the knowledge base.", |
| "cite_spans": [ |
| { |
| "start": 104, |
| "end": 121, |
| "text": "Lin et al. (2018)", |
| "ref_id": "BIBREF15" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Related Work", |
| "sec_num": "2" |
| }, |
| { |
| "text": "With respect to WordNet, this work is related to another research direction, that of automatically creating sense-level lexicons starting from wordlevel lexicons. Two commonly used such lexicons are SentiWordNet (Baccianella et al., 2010) and +/-EffectWordNet (Choi and Wiebe, 2014) . The first one is a sense-level lexicon providing WordNet synsets with prior sentiment polarity annotations. +/-EffectWordNet provides annotations for synsets that express negative or positive effects over entities. Our work of annotating WordNet synsets as well as other KG concepts with scores reflecting their relevance to moral values is complementary to the previously mentioned lexicons, as they bring in the dimension of morality.", |
| "cite_spans": [ |
| { |
| "start": 212, |
| "end": 238, |
| "text": "(Baccianella et al., 2010)", |
| "ref_id": "BIBREF2" |
| }, |
| { |
| "start": 260, |
| "end": 282, |
| "text": "(Choi and Wiebe, 2014)", |
| "ref_id": "BIBREF4" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Related Work", |
| "sec_num": "2" |
| }, |
| { |
| "text": "In our work, we investigate how moral values are captured in KGs such as DBpedia, WordNet, and ConceptNet. We use the same intuition across all three KGs: if terms in the MFD can be linked to corresponding concepts in the KG, then the semantic relations contained in the KGs can be exploited to score all the other concepts with respect to their relevance to the moral values. These relevance scores can subsequently be used, for example, for expanding the lexicon. Alternatively, they can be used directly as features in applications aiming at classifying texts based on moral values.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Related Work", |
| "sec_num": "2" |
| }, |
| { |
| "text": "Another important motivation for our work is that KGs have the benefit of providing structured knowledge, but most of the time, this knowledge is factual (e.g., DBpedia) or lexical (e.g., WordNet). Most KGs, ConceptNet being one exception, lack in common sense knowledge. And knowledge such as what is generally accepted as morally good or bad is also missing in ConceptNet. This work is a step in the direction of enriching KGs with such common-sense knowledge.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Related Work", |
| "sec_num": "2" |
| }, |
| { |
| "text": "The core idea of our approach is to map the MFD lexicon to a KG, and subsequently score all entities in the KG with respect to their relevance to moral values. We are taking a layered approach to the moral foundations, as we are interested in scoring the relevance on three levels: (1) the moral trait level, in which we are interested in the relevance of concepts with respect to each moral trait (virtue / vice). Hence, each concept obtains 10 scores, two (virtue and vice) for each moral foundation.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Approach", |
| "sec_num": "3" |
| }, |
| { |
| "text": "(2) the moral foundation level that scores each concept with respect to the five moral foundations; and (3) the the moral polarity level that scores each concept with respect to its relevance to vices and virtues.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Approach", |
| "sec_num": "3" |
| }, |
| { |
| "text": "As a measure of relevance in KGs, we use Personalized PageRank (Haveliwala, 2003 )(PPR). Our work consists of two main steps: first, we manually link the MFD entries to entries in the KGs. Then, we use these KG entries as seeds for running PPR. We now describe both steps in more detail.", |
| "cite_spans": [ |
| { |
| "start": 63, |
| "end": 80, |
| "text": "(Haveliwala, 2003", |
| "ref_id": "BIBREF12" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Approach", |
| "sec_num": "3" |
| }, |
| { |
| "text": "We manually link each entry in the lexicon to the corresponding concept(s) in the KGs. This step involves disambiguation judgments. Also, multiple concepts can be linked to the same lexicon entry, as long as they are related to the corresponding moral value.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Linking the MFD to KGs", |
| "sec_num": "3.1" |
| }, |
| { |
| "text": "Linking to WordNet 3.1 (WN) The concepts in WN which we focus on are the synsets. As each synset has a specific meaning, we aim to only link to concepts that are truly relevant for the respective moral foundation. For example, the word fair occurs as a synonym for just which is relevant for the moral foundation of fairness/Cheating, but also as a gathering of producers to promote business. Thus, we manually decided for each synset, which contains an entry of the MFD lexicon, whether it relates to the moral foundation or not.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Linking the MFD to KGs", |
| "sec_num": "3.1" |
| }, |
| { |
| "text": "Linking to ConceptNet (CN) In CN, the concepts are only disambiguated with respect to their part of speech. As this kind of disambiguation is not needed for linking the entries and in order to obtain a graph that is less sparse, we collapsed all concepts that only differ in their part of speech. Further, we remove so-called External Concepts, which are links to other resources such as WordNet as well as all isolated nodes. For classifying the specific moral traits and the moral polarity, we remove such relations between concepts that express a semantic difference: Antonym, DistinctFrom, Not-CapableOf, NotDesires, NotHasProperty. For linking to ConceptNet, we check which single word concepts match the lexicon entries and manually verify these links to avoid nonsense (i.e., Churchill matches church*, but we do not expect it to be an intended cue word for purity/sanctity). As this way of linking often includes inflected forms of words (i.e. care* also includes cares, cared, caring and even careth), we further ignore DerivedFrom and FormOf relations in our leave-one-out evaluation.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Linking the MFD to KGs", |
| "sec_num": "3.1" |
| }, |
| { |
| "text": "Linking to DBpedia For linking the lexicon entries to concepts in DBpedia, we use the following process: we check in Wikipedia for articles whose name is the lexicon entry. If such an article exists, it is related to the moral value, and it is not a disambiguation page, then we add the link to the sense lexicon. If the formed URL is redirected, we check if the redirected article is related to the moral value. If so, then we add it to the sense lexicon. If an article or a redirected page does not exist, but a disambiguation page, then we check each disambiguation article listed in the disambiguation page and select the ones that are related to the moral trait. This is the case, for instance, for the term shelter with Wikipedia disambiguation page www.wikipedia.org/wiki/Shelter. This page provides many disambiguation options for the term, including Shelter (building), Animal shelter, Homeless shelter which we add to the lexicon, but also others that we deem unrelated to the intended meaning of the MFD, for example, multiple locations such as Port Shelter -a harbor in Hong Kong, multiple films, albums, singles, novels, video games, and others that we do not add to the lexicon. In Wikipedia, it is often the case that an article is the default for a particular word, but that a Wikipedia disambiguation page also exists for the same word. If that is the case, we add the default article if it is related to the moral trait, and also check the disambiguation page, and manually select from all the disambiguation articles the ones that are related to the moral trait. We then map all the collected Wikipedia articles to their corresponding DBpedia resources.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Linking the MFD to KGs", |
| "sec_num": "3.1" |
| }, |
| { |
| "text": "The exact number of concepts that we obtain in this process for each moral trait and each KG are shown in Table 1 . We use these concepts as seeds for computing relevance scores of all the concepts in the KG, as described in the following section.", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 106, |
| "end": 113, |
| "text": "Table 1", |
| "ref_id": "TABREF1" |
| } |
| ], |
| "eq_spans": [], |
| "section": "Linking the MFD to KGs", |
| "sec_num": "3.1" |
| }, |
| { |
| "text": "We use Personalized PageRank as a measure of relevance in KGs. Relations in KGs are typed and directed. However, the relations are semantic. Therefore for each relation, one can consider that another relation in the opposite direction also exists. For example, for each occupationOf relation, the hasOccupation relation can be defined, pointing in the opposite direction. Since PPR is working on directed networks, but we want the random walker to be able to also follow incoming edges, we add for each relation with type t in the KGs, an additional, opposite relation whose type we set to inv t. We investigate 3 ways of computing the link probabilities in PPR: Uniform (U) which is the standard PPR, disregarding the edge types; Type-Uniform (TU): the random walker chooses uniformly at random one of the available edge types; Then, for the chosen type, it chooses uniformly at random one of the edges with that particular type; Type exclusivity (TE): the random walker first chooses an edge type uniformly at random. Then, among the edges of that type, it chooses which one to take according to the exclusivity of the edge. Exclusivity (Hulpu\u015f et al., 2015 ) is a measure of relation importance that provides higher scores to relation types with low cardinality for both source and target nodes.", |
| "cite_spans": [ |
| { |
| "start": 1139, |
| "end": 1159, |
| "text": "(Hulpu\u015f et al., 2015", |
| "ref_id": "BIBREF13" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Personalized PageRank for Moral Value Relevance Scoring", |
| "sec_num": "3.2" |
| }, |
| { |
| "text": "On the moral trait level, we consider each of the 10 classes of traits in the MFD individually, excluding the general class whose entries are not split according to their polarity. For each trait, we run a PPR process where the teleport probability is distributed uniformly to the seeds of the trait. Consequently, each concept in the KG receives 10 On the moral foundation level, we score all concepts with respect to their relevance to the five moral foundations. We create the set of seeds for each foundation by merging the corresponding sets of vice and virtue seeds. One PPR process is run for each foundation, as well as on the general morality class, therefore each concept in the KG receives 6 scores for each PPR method. As previously, the teleport probability of the PPR process is shared uniformly by the seeds.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Personalized PageRank for Moral Value Relevance Scoring", |
| "sec_num": "3.2" |
| }, |
| { |
| "text": "On the moral polarity level, we create the set of seeds for each of the two classes (vices and virtues) by merging the vice seeds and the virtue seeds of all foundations, respectively. Therefore, we provide each concept with two scores for each PPR method.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Personalized PageRank for Moral Value Relevance Scoring", |
| "sec_num": "3.2" |
| }, |
| { |
| "text": "We evaluate the prediction on each level independently. Similarly to Xie et al. (2019), we use a leave-one-out evaluation. Specifically, for each seed concept of a class, we run an additional PPR process for that class when the targeted seed concept is left out of the teleport vector. Then, we check the relevance score the targeted seed concept obtains in this PPR process and compare it to its relevance scores for the other classes.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Experiments and Discussion", |
| "sec_num": "4" |
| }, |
| { |
| "text": "To measure the accuracy of the prediction, we take a very straightforward approach and consider a hit when the targeted seed concept achieves the maximum relevance score for the class to which it belongs when it is left out. Therefore, we compute the accuracy for a class c as the percent of seed concepts of c that obtained the maximum score across all classes for class c, when their score for class c is computed as their PPR score when left out of the seed set. Table 2 shows the results obtained for each PPR version, per trait as well as overall, for each of the three KGs. Since there are 10 classes, a random baseline assignment would obtain, on average, a .10 score. Therefore, all methods manage to perform substantially better than random on all KGs.", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 466, |
| "end": 473, |
| "text": "Table 2", |
| "ref_id": "TABREF3" |
| } |
| ], |
| "eq_spans": [], |
| "section": "Experiments and Discussion", |
| "sec_num": "4" |
| }, |
| { |
| "text": "Regarding the methods, we observe that the scores differ between the classes, particularly for DBpedia. For example, the Uniform PPR achieves a .79 score for class Fairness-vice on DBpedia, the Type Uniform PPR achieves .71, and the Type-Exclusivity .67. The class Authority-vice is also handled very differently by the three methods on DBpedia. On the other KGs, the different PPR methods do not show significant differences, with their performances being most similar on Concept-Net. Indeed, DBpedia provides many relation types with varied cardinality, so it is not surprising that methods that treat relation types differently obtain significantly different results on this KG. However, interestingly, the overall results of all the three methods are very similar, including on DBpedia.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Results of Moral Trait Prediction", |
| "sec_num": "4.1" |
| }, |
| { |
| "text": "With respect to KGs, ConceptNet is in a strong lead over both WordNet and DBpedia, with all methods obtaining scores between .72 and .98 on the individual moral traits. All methods perform worse on WordNet with .59 overall scores. These values are similar to the highest scores obtained by Xie et al. (2019) when running their leave-oneout classification of moral traits, specifically where the MFD entries are embedded using the Google N-grams corpus (Lin et al., 2012) and the classification is done with a Centroid model. DBpedia captures the moral traits the worst among the three KGs, with a wide range of values across different classes and an overall score of .43 or .44 depending on the method. This perfor- Regarding the different moral traits, several scores stand out, particularly the almost random performance of all methods on DBpedia for the Loyalty-vice class. As seen in Table 1 , we managed to identify only 15 DBpedia concepts for this class. Among them, many are also present in the set of concepts of other moral traits. For instance, dbpedia.org/resource/Apostasy also belongs to the Authority-vice and to the Purity-vice classes, while three concepts related to MFD entry abandon are also part of the seed concepts of Carevice. Among the MFD entries that only occur in this class, for many, we did not find a corresponding concept in DBpedia, for instance, imposter, jilt*, miscreant, renegate. The confusion matrix shown in Table 4 reveals that more entries of the Loyaltyvice class achieve higher relevance with respect to the Authority-vice and Care-vice traits rather than with respect to the Loyalty vice trait. Also on WordNet, Loyalty-vice prediction is relatively often mistaken for Authority-vice and Care-vice, as seen in the confusion matrix of Table 3 .", |
| "cite_spans": [ |
| { |
| "start": 452, |
| "end": 470, |
| "text": "(Lin et al., 2012)", |
| "ref_id": "BIBREF16" |
| } |
| ], |
| "ref_spans": [ |
| { |
| "start": 888, |
| "end": 895, |
| "text": "Table 1", |
| "ref_id": "TABREF1" |
| }, |
| { |
| "start": 1448, |
| "end": 1455, |
| "text": "Table 4", |
| "ref_id": "TABREF6" |
| }, |
| { |
| "start": 1779, |
| "end": 1786, |
| "text": "Table 3", |
| "ref_id": "TABREF5" |
| } |
| ], |
| "eq_spans": [], |
| "section": "Results of Moral Trait Prediction", |
| "sec_num": "4.1" |
| }, |
| { |
| "text": "As seen in Tables 3 and 4 , in both WordNet and DBpedia, vices are usually confused with vices 2 https://www.english-corpora.org/coha/ of another foundation, and virtues with virtues of other foundations, and when that is not the case, then a vice is confused with a virtue of the same dimension and the other way round. An exception is the prediction of Loyalty-virtue on DBpedia, which is often mistaken for Fairness-Vice. This is likely due to their shared MFD lexicon entry segregation, for which we encounter 7 concepts in DBpedia. Table 5 shows the results obtained on the moral dimension level, for each PPR version, per class as well as overall, for each of the three KGs. Since we consider 6 classes, a random baseline achieves on average .17, therefore again, all methods on all KGs significantly outperform this trivial baseline.", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 11, |
| "end": 25, |
| "text": "Tables 3 and 4", |
| "ref_id": "TABREF5" |
| }, |
| { |
| "start": 537, |
| "end": 544, |
| "text": "Table 5", |
| "ref_id": "TABREF8" |
| } |
| ], |
| "eq_spans": [], |
| "section": "Results of Moral Trait Prediction", |
| "sec_num": "4.1" |
| }, |
| { |
| "text": "The Overall results come to reinforce our conclusion from the previous analysis that the treatment of relation types in the PPR process is only beneficial for DBpedia. As expected, the results of this prediction are better than of the moral-trait prediction, and this improvement is particularly strong on DBpedia.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Results of Moral Foundation Prediction", |
| "sec_num": "4.2" |
| }, |
| { |
| "text": "Among the foundations, the Authority/Subversion foundation achieves under average scores on all methods on all KGs, while the Care/Harm foundation is correctly predicted more often than average by all methods on all KGs. The ConceptNet prediction of the Fairness/Cheating dimension stands out through its very high scores.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Results of Moral Foundation Prediction", |
| "sec_num": "4.2" |
| }, |
| { |
| "text": "The prediction of the Loyalty/Betrayal foundation stands out for its poor scores on DBpedia. On the other side, the Purity/Degradation foundation is the foundation predicted best on DBpedia, achieving a score even higher than WordNet's prediction. Also the general morality class is captured quite well by all KGs, including on DBpedia, where its prediction is more accurate than on WordNet.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Results of Moral Foundation Prediction", |
| "sec_num": "4.2" |
| }, |
| { |
| "text": "In Table 6 , we present the confusion matrix of predicting the foundations for DBpedia and Word-Net. As previously, since ConceptNet has very high scores, the confusion matrix is not conclusive, so we do not report it. Interestingly, in DBpedia, only Purity/Degradation is not mostly confused with Fairness/Cheating. Table 7 shows the results obtained on the moral polarity level, for each PPR version, per class as well as overall, for each of the three KGs. For this prediction, the random baseline would achieve, on average, a .5 score. Again, all methods achieve for all KGs significantly higher scores than the random baseline. On DBpedia, the prediction of virtues achieves slightly higher scores than the prediction of vices, while on WordNet and ConceptNet, the opposite holds. In ConceptNet, the prediction of vices vs. virtues achieves very high scores of .96 and .97, respectively. For comparison, Xie et al. (2019) report .93 accuracy on predicting the polarity when using Google N-grams embeddings and a 5-NN model. With COHA embeddings, the highest accuracy is .80, obtained with the Centroid model.", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 3, |
| "end": 10, |
| "text": "Table 6", |
| "ref_id": "TABREF9" |
| }, |
| { |
| "start": 317, |
| "end": 324, |
| "text": "Table 7", |
| "ref_id": "TABREF11" |
| } |
| ], |
| "eq_spans": [], |
| "section": "Results of Moral Foundation Prediction", |
| "sec_num": "4.2" |
| }, |
| { |
| "text": "To also give an intuition of how our approach scores concepts that are not seeds, we also report the top 10 highest scored concepts that are not seeds, for the fairness/cheating and the authority/subversion foundations in Table 8 .", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 222, |
| "end": 229, |
| "text": "Table 8", |
| "ref_id": "TABREF12" |
| } |
| ], |
| "eq_spans": [], |
| "section": "Qualitative Analysis", |
| "sec_num": "4.4" |
| }, |
| { |
| "text": "WordNet concepts found for fairness/cheating are quite reasonable, while for subversion, there are some false positives (jurisprudence, loyalty). In ConceptNet, we overall find similar concepts for fairness/cheating, while some are on the wrong side, such as bias, prejudice, fair, equal, judge. In the authority/subversion foundation, often, the same concepts are scored high for both the vice and virtue (us), while some that are rather on the wrong side (detraditionalize, obedience, law).", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Qualitative Analysis", |
| "sec_num": "4.4" |
| }, |
| { |
| "text": "DBpedia ranks high entities that are much different from those of WordNet and ConceptNet. While the high scores of some entities are not easily understandable (i.e. National debt of the United States, Frank Stanford), others nicely capture some background knowledge about the entities: i.e., Coretta Scott King and Zechariah Chafee both espoused civil rights. Interestingly, just as for ConceptNet, the United States is scored as highly relevant for multiple traits. This is because many DBpedia concepts that we linked to the MFD through the disambiguation pages are related to the United States, for instance, segregation (Fairness-vice) has been linked among others to Housing segregation in the United States and Residential segregation in the United States.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Qualitative Analysis", |
| "sec_num": "4.4" |
| }, |
| { |
| "text": "In this work, we investigated how moral traits, foundations, and polarity based on the MFT are captured in three widely used KGs. Our analysis reveals big differences between the three explored KGs, both quantitatively and qualitatively. Con-ceptNet achieves high accuracies at predicting the class of seeds in a leave-one-out evaluation. The results on WordNet are well aligned with results obtained by related work in a similar evaluation. Lastly, the seed class prediction accuracy of DBpedia scores last among the three datasets, but still significantly higher than random, and it comes with the advantage of dealing with entities such as people and organizations.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Conclusion", |
| "sec_num": "5" |
| }, |
| { |
| "text": "All KGs manage to accurately discriminate between virtues and vices, which is already a great step towards automatically telling the good from the bad. The more complex problem of predicting the foundation and the granular trait can still undergo substantial improvements, particularly on WordNet and DBpedia. However, given that our method is completely unsupervised, using just Personalized PageRank, we conclude that there is great potential in bringing morality common-sense into knowledge graphs. As future work, we are committed to further analyzing our approach, particularly on applications such as classification of texts with respect to moral values.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Conclusion", |
| "sec_num": "5" |
| }, |
| { |
| "text": "All resources created in this work, as well as all relevance scores for the KGs, are available at https://github. com/dwslab/Morality-in-Knowledge-Graphs.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| } |
| ], |
| "back_matter": [ |
| { |
| "text": "This work has been funded by the Deutsche Forschungsgemeinschaft (DFG) within the project ExpLAIN, Grant Number STU 266/14-1, as part of the Priority Program \"Robust Argumentation Machines (RATIO)\" (SPP-1999), as well as by the Natural Sciences and Engineering Research Council of Canada.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Acknowledgments", |
| "sec_num": null |
| } |
| ], |
| "bib_entries": { |
| "BIBREF0": { |
| "ref_id": "b0", |
| "title": "Association of moral values with vaccine hesitancy", |
| "authors": [ |
| { |
| "first": "Robert", |
| "middle": [ |
| "A" |
| ], |
| "last": "Avnika B Amin", |
| "suffix": "" |
| }, |
| { |
| "first": "Cara", |
| "middle": [ |
| "E" |
| ], |
| "last": "Bednarczyk", |
| "suffix": "" |
| }, |
| { |
| "first": "Kala", |
| "middle": [ |
| "J" |
| ], |
| "last": "Ray", |
| "suffix": "" |
| }, |
| { |
| "first": "Jesse", |
| "middle": [], |
| "last": "Melchiori", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Graham", |
| "suffix": "" |
| }, |
| { |
| "first": "R", |
| "middle": [], |
| "last": "Jeffrey", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Huntsinger", |
| "suffix": "" |
| }, |
| { |
| "first": "B", |
| "middle": [], |
| "last": "Saad", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Omer", |
| "suffix": "" |
| } |
| ], |
| "year": 2017, |
| "venue": "Nature Human Behaviour", |
| "volume": "1", |
| "issue": "12", |
| "pages": "873--880", |
| "other_ids": { |
| "DOI": [ |
| "10.1038/s41562-017-0256-5" |
| ] |
| }, |
| "num": null, |
| "urls": [], |
| "raw_text": "Avnika B Amin, Robert A Bednarczyk, Cara E Ray, Kala J Melchiori, Jesse Graham, Jeffrey R Huntsinger, and Saad B Omer. 2017. Association of moral values with vaccine hesitancy. Nature Human Behaviour, 1(12):873-880.", |
| "links": null |
| }, |
| "BIBREF1": { |
| "ref_id": "b1", |
| "title": "Moralstrength: Exploiting a moral lexicon and embedding similarity for moral foundations prediction. Knowledge-Based Systems", |
| "authors": [ |
| { |
| "first": "Oscar", |
| "middle": [], |
| "last": "Araque", |
| "suffix": "" |
| }, |
| { |
| "first": "Lorenzo", |
| "middle": [], |
| "last": "Gatti", |
| "suffix": "" |
| }, |
| { |
| "first": "Kyriaki", |
| "middle": [], |
| "last": "Kalimeri", |
| "suffix": "" |
| } |
| ], |
| "year": 2020, |
| "venue": "", |
| "volume": "191", |
| "issue": "", |
| "pages": "", |
| "other_ids": { |
| "DOI": [ |
| "10.1016/j.knosys.2019.105184" |
| ] |
| }, |
| "num": null, |
| "urls": [], |
| "raw_text": "Oscar Araque, Lorenzo Gatti, and Kyriaki Kalimeri. 2020. Moralstrength: Exploiting a moral lexicon and embedding similarity for moral foundations pre- diction. Knowledge-Based Systems, 191:105184.", |
| "links": null |
| }, |
| "BIBREF2": { |
| "ref_id": "b2", |
| "title": "SentiWordNet 3.0: An enhanced lexical resource for sentiment analysis and opinion mining", |
| "authors": [ |
| { |
| "first": "Stefano", |
| "middle": [], |
| "last": "Baccianella", |
| "suffix": "" |
| }, |
| { |
| "first": "Andrea", |
| "middle": [], |
| "last": "Esuli", |
| "suffix": "" |
| }, |
| { |
| "first": "Fabrizio", |
| "middle": [], |
| "last": "Sebastiani", |
| "suffix": "" |
| } |
| ], |
| "year": 2010, |
| "venue": "Proceedings of the Seventh International Conference on Language Resources and Evaluation (LREC'10)", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Stefano Baccianella, Andrea Esuli, and Fabrizio Sebas- tiani. 2010. SentiWordNet 3.0: An enhanced lexi- cal resource for sentiment analysis and opinion min- ing. In Proceedings of the Seventh International Conference on Language Resources and Evaluation (LREC'10), Valletta, Malta. European Language Re- sources Association (ELRA).", |
| "links": null |
| }, |
| "BIBREF3": { |
| "ref_id": "b3", |
| "title": "The morality of may 2, 2011: A content analysis of us headlines regarding the death of osama bin laden", |
| "authors": [ |
| { |
| "first": "Nicholas", |
| "middle": [], |
| "last": "Bowman", |
| "suffix": "" |
| }, |
| { |
| "first": "Robert", |
| "middle": [ |
| "Joel" |
| ], |
| "last": "Lewis", |
| "suffix": "" |
| }, |
| { |
| "first": "Ron", |
| "middle": [], |
| "last": "Tamborini", |
| "suffix": "" |
| } |
| ], |
| "year": 2014, |
| "venue": "", |
| "volume": "17", |
| "issue": "", |
| "pages": "639--664", |
| "other_ids": { |
| "DOI": [ |
| "10.1080/15205436.2013.822518" |
| ] |
| }, |
| "num": null, |
| "urls": [], |
| "raw_text": "Nicholas Bowman, Robert Joel Lewis, and Ron Tam- borini. 2014. The morality of may 2, 2011: A con- tent analysis of us headlines regarding the death of osama bin laden. Mass Communication and Society, 17(5):639-664.", |
| "links": null |
| }, |
| "BIBREF4": { |
| "ref_id": "b4", |
| "title": "+/-EffectWordNet: Sense-level lexicon acquisition for opinion inference", |
| "authors": [ |
| { |
| "first": "Yoonjung", |
| "middle": [], |
| "last": "Choi", |
| "suffix": "" |
| }, |
| { |
| "first": "Janyce", |
| "middle": [], |
| "last": "Wiebe", |
| "suffix": "" |
| } |
| ], |
| "year": 2014, |
| "venue": "Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP)", |
| "volume": "", |
| "issue": "", |
| "pages": "1181--1191", |
| "other_ids": { |
| "DOI": [ |
| "10.3115/v1/D14-1125" |
| ] |
| }, |
| "num": null, |
| "urls": [], |
| "raw_text": "Yoonjung Choi and Janyce Wiebe. 2014. +/- EffectWordNet: Sense-level lexicon acquisition for opinion inference. In Proceedings of the 2014 Con- ference on Empirical Methods in Natural Language Processing (EMNLP), pages 1181-1191, Doha, Qatar. Association for Computational Linguistics.", |
| "links": null |
| }, |
| "BIBREF5": { |
| "ref_id": "b5", |
| "title": "How words do the work of politics: Moral foundations theory and the debate over stem cell research", |
| "authors": [ |
| { |
| "first": "Scott", |
| "middle": [], |
| "last": "Clifford", |
| "suffix": "" |
| }, |
| { |
| "first": "Jennifer", |
| "middle": [], |
| "last": "Jerit", |
| "suffix": "" |
| } |
| ], |
| "year": 2013, |
| "venue": "The Journal of Politics", |
| "volume": "75", |
| "issue": "3", |
| "pages": "659--671", |
| "other_ids": { |
| "DOI": [ |
| "http://www.jstor.org/stable/10.1017/s0022381613000492" |
| ] |
| }, |
| "num": null, |
| "urls": [], |
| "raw_text": "Scott Clifford and Jennifer Jerit. 2013. How words do the work of politics: Moral foundations theory and the debate over stem cell research. The Journal of Politics, 75(3):659-671.", |
| "links": null |
| }, |
| "BIBREF6": { |
| "ref_id": "b6", |
| "title": "Journal of the American society for information science", |
| "authors": [ |
| { |
| "first": "Scott", |
| "middle": [], |
| "last": "Deerwester", |
| "suffix": "" |
| }, |
| { |
| "first": "T", |
| "middle": [], |
| "last": "Susan", |
| "suffix": "" |
| }, |
| { |
| "first": "George", |
| "middle": [ |
| "W" |
| ], |
| "last": "Dumais", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Furnas", |
| "suffix": "" |
| }, |
| { |
| "first": "K", |
| "middle": [], |
| "last": "Thomas", |
| "suffix": "" |
| }, |
| { |
| "first": "Richard", |
| "middle": [], |
| "last": "Landauer", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Harshman", |
| "suffix": "" |
| } |
| ], |
| "year": 1990, |
| "venue": "", |
| "volume": "41", |
| "issue": "", |
| "pages": "391--407", |
| "other_ids": { |
| "DOI": [ |
| "10.1002/(SICI)1097-4571(199009)41:6<391::AID-ASI1>3.0.CO;2-9" |
| ] |
| }, |
| "num": null, |
| "urls": [], |
| "raw_text": "Scott Deerwester, Susan T Dumais, George W Fur- nas, Thomas K Landauer, and Richard Harshman. 1990. Indexing by latent semantic analysis. Jour- nal of the American society for information science, 41(6):391-407.", |
| "links": null |
| }, |
| "BIBREF7": { |
| "ref_id": "b7", |
| "title": "Purity homophily in social networks", |
| "authors": [ |
| { |
| "first": "Morteza", |
| "middle": [], |
| "last": "Dehghani", |
| "suffix": "" |
| }, |
| { |
| "first": "Kate", |
| "middle": [], |
| "last": "Johnson", |
| "suffix": "" |
| }, |
| { |
| "first": "Joe", |
| "middle": [], |
| "last": "Hoover", |
| "suffix": "" |
| }, |
| { |
| "first": "Eyal", |
| "middle": [], |
| "last": "Sagi", |
| "suffix": "" |
| }, |
| { |
| "first": "Justin", |
| "middle": [], |
| "last": "Garten", |
| "suffix": "" |
| }, |
| { |
| "first": "Niki", |
| "middle": [ |
| "Jitendra" |
| ], |
| "last": "Parmar", |
| "suffix": "" |
| }, |
| { |
| "first": "Stephen", |
| "middle": [], |
| "last": "Vaisey", |
| "suffix": "" |
| }, |
| { |
| "first": "Rumen", |
| "middle": [], |
| "last": "Iliev", |
| "suffix": "" |
| }, |
| { |
| "first": "Jesse", |
| "middle": [], |
| "last": "Graham", |
| "suffix": "" |
| } |
| ], |
| "year": 2016, |
| "venue": "Journal of Experimental Psychology: General", |
| "volume": "145", |
| "issue": "3", |
| "pages": "", |
| "other_ids": { |
| "DOI": [ |
| "10.1037/xge0000139" |
| ] |
| }, |
| "num": null, |
| "urls": [], |
| "raw_text": "Morteza Dehghani, Kate Johnson, Joe Hoover, Eyal Sagi, Justin Garten, Niki Jitendra Parmar, Stephen Vaisey, Rumen Iliev, and Jesse Graham. 2016. Pu- rity homophily in social networks. Journal of Ex- perimental Psychology: General, 145(3):366.", |
| "links": null |
| }, |
| "BIBREF8": { |
| "ref_id": "b8", |
| "title": "Morality between the lines: Detecting moral sentiment in text", |
| "authors": [ |
| { |
| "first": "Justin", |
| "middle": [], |
| "last": "Garten", |
| "suffix": "" |
| }, |
| { |
| "first": "Reihane", |
| "middle": [], |
| "last": "Boghrati", |
| "suffix": "" |
| }, |
| { |
| "first": "Joe", |
| "middle": [], |
| "last": "Hoover", |
| "suffix": "" |
| }, |
| { |
| "first": "Kate", |
| "middle": [ |
| "M" |
| ], |
| "last": "Johnson", |
| "suffix": "" |
| }, |
| { |
| "first": "Morteza", |
| "middle": [], |
| "last": "Dehghani", |
| "suffix": "" |
| } |
| ], |
| "year": 2016, |
| "venue": "Proceedings of IJCAI 2016 workshop on Computational Modeling of Attitudes", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Justin Garten, Reihane Boghrati, Joe Hoover, Kate M Johnson, and Morteza Dehghani. 2016. Morality be- tween the lines: Detecting moral sentiment in text. In Proceedings of IJCAI 2016 workshop on Compu- tational Modeling of Attitudes.", |
| "links": null |
| }, |
| "BIBREF9": { |
| "ref_id": "b9", |
| "title": "Moral foundations theory: The pragmatic validity of moral pluralism", |
| "authors": [ |
| { |
| "first": "Jesse", |
| "middle": [], |
| "last": "Graham", |
| "suffix": "" |
| }, |
| { |
| "first": "Jonathan", |
| "middle": [], |
| "last": "Haidt", |
| "suffix": "" |
| }, |
| { |
| "first": "Sena", |
| "middle": [], |
| "last": "Koleva", |
| "suffix": "" |
| }, |
| { |
| "first": "Matt", |
| "middle": [], |
| "last": "Motyl", |
| "suffix": "" |
| }, |
| { |
| "first": "Ravi", |
| "middle": [], |
| "last": "Iyer", |
| "suffix": "" |
| }, |
| { |
| "first": "P", |
| "middle": [], |
| "last": "Sean", |
| "suffix": "" |
| }, |
| { |
| "first": "Peter", |
| "middle": [ |
| "H" |
| ], |
| "last": "Wojcik", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Ditto", |
| "suffix": "" |
| } |
| ], |
| "year": 2013, |
| "venue": "Advances in experimental social psychology", |
| "volume": "47", |
| "issue": "", |
| "pages": "55--130", |
| "other_ids": { |
| "DOI": [ |
| "10.1016/B978-0-12-407236-7.00002-4" |
| ] |
| }, |
| "num": null, |
| "urls": [], |
| "raw_text": "Jesse Graham, Jonathan Haidt, Sena Koleva, Matt Motyl, Ravi Iyer, Sean P Wojcik, and Peter H Ditto. 2013. Moral foundations theory: The pragmatic va- lidity of moral pluralism. In Advances in experimen- tal social psychology, volume 47, pages 55-130. El- sevier.", |
| "links": null |
| }, |
| "BIBREF10": { |
| "ref_id": "b10", |
| "title": "Liberals and conservatives rely on different sets of moral foundations", |
| "authors": [ |
| { |
| "first": "Jesse", |
| "middle": [], |
| "last": "Graham", |
| "suffix": "" |
| }, |
| { |
| "first": "Jonathan", |
| "middle": [], |
| "last": "Haidt", |
| "suffix": "" |
| }, |
| { |
| "first": "Brian", |
| "middle": [ |
| "A" |
| ], |
| "last": "Nosek", |
| "suffix": "" |
| } |
| ], |
| "year": 2009, |
| "venue": "Journal of personality and social psychology", |
| "volume": "96", |
| "issue": "5", |
| "pages": "", |
| "other_ids": { |
| "DOI": [ |
| "10.1037/a0015141" |
| ] |
| }, |
| "num": null, |
| "urls": [], |
| "raw_text": "Jesse Graham, Jonathan Haidt, and Brian A Nosek. 2009. Liberals and conservatives rely on different sets of moral foundations. Journal of personality and social psychology, 96(5):1029.", |
| "links": null |
| }, |
| "BIBREF11": { |
| "ref_id": "b11", |
| "title": "Intuitive ethics: How innately prepared intuitions generate culturally variable virtues", |
| "authors": [ |
| { |
| "first": "Jonathan", |
| "middle": [], |
| "last": "Haidt", |
| "suffix": "" |
| }, |
| { |
| "first": "Craig", |
| "middle": [], |
| "last": "Joseph", |
| "suffix": "" |
| } |
| ], |
| "year": 2004, |
| "venue": "Daedalus", |
| "volume": "133", |
| "issue": "4", |
| "pages": "55--66", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Jonathan Haidt and Craig Joseph. 2004. Intuitive ethics: How innately prepared intuitions generate culturally variable virtues. Daedalus, 133(4):55-66.", |
| "links": null |
| }, |
| "BIBREF12": { |
| "ref_id": "b12", |
| "title": "Topic-sensitive pagerank: a context-sensitive ranking algorithm for web search", |
| "authors": [ |
| { |
| "first": "T", |
| "middle": [ |
| "H" |
| ], |
| "last": "Haveliwala", |
| "suffix": "" |
| } |
| ], |
| "year": 2003, |
| "venue": "IEEE Transactions on Knowledge and Data Engineering", |
| "volume": "15", |
| "issue": "4", |
| "pages": "784--796", |
| "other_ids": { |
| "DOI": [ |
| "10.1109/TKDE.2003.1208999" |
| ] |
| }, |
| "num": null, |
| "urls": [], |
| "raw_text": "T. H. Haveliwala. 2003. Topic-sensitive pagerank: a context-sensitive ranking algorithm for web search. IEEE Transactions on Knowledge and Data Engi- neering, 15(4):784-796.", |
| "links": null |
| }, |
| "BIBREF13": { |
| "ref_id": "b13", |
| "title": "Path-based semantic relatedness on linked data and its use to word and entity disambiguation", |
| "authors": [ |
| { |
| "first": "Ioana", |
| "middle": [], |
| "last": "Hulpu\u015f", |
| "suffix": "" |
| }, |
| { |
| "first": "Narumol", |
| "middle": [], |
| "last": "Prangnawarat", |
| "suffix": "" |
| }, |
| { |
| "first": "Conor", |
| "middle": [], |
| "last": "Hayes", |
| "suffix": "" |
| } |
| ], |
| "year": 2015, |
| "venue": "The Semantic Web -ISWC 2015", |
| "volume": "", |
| "issue": "", |
| "pages": "442--457", |
| "other_ids": { |
| "DOI": [ |
| "10.1007/978-3-319-25007-6_26" |
| ] |
| }, |
| "num": null, |
| "urls": [], |
| "raw_text": "Ioana Hulpu\u015f, Narumol Prangnawarat, and Conor Hayes. 2015. Path-based semantic relatedness on linked data and its use to word and entity disam- biguation. In The Semantic Web -ISWC 2015, pages 442-457, Cham. Springer International Publishing.", |
| "links": null |
| }, |
| "BIBREF14": { |
| "ref_id": "b14", |
| "title": "Quantifying moral foundations from various topics on twitter conversations", |
| "authors": [ |
| { |
| "first": "Rishemjit", |
| "middle": [], |
| "last": "Kaur", |
| "suffix": "" |
| }, |
| { |
| "first": "Kazutoshi", |
| "middle": [], |
| "last": "Sasahara", |
| "suffix": "" |
| } |
| ], |
| "year": 2016, |
| "venue": "2016 IEEE International Conference on Big Data (Big Data)", |
| "volume": "", |
| "issue": "", |
| "pages": "2505--2512", |
| "other_ids": { |
| "DOI": [ |
| "10.1109/BigData.2016.7840889" |
| ] |
| }, |
| "num": null, |
| "urls": [], |
| "raw_text": "Rishemjit Kaur and Kazutoshi Sasahara. 2016. Quanti- fying moral foundations from various topics on twit- ter conversations. In 2016 IEEE International Con- ference on Big Data (Big Data), pages 2505-2512.", |
| "links": null |
| }, |
| "BIBREF15": { |
| "ref_id": "b15", |
| "title": "Acquiring background knowledge to improve moral value prediction", |
| "authors": [ |
| { |
| "first": "Ying", |
| "middle": [], |
| "last": "Lin", |
| "suffix": "" |
| }, |
| { |
| "first": "Joe", |
| "middle": [], |
| "last": "Hoover", |
| "suffix": "" |
| }, |
| { |
| "first": "Gwenyth", |
| "middle": [], |
| "last": "Portillo-Wightman", |
| "suffix": "" |
| }, |
| { |
| "first": "Christina", |
| "middle": [], |
| "last": "Park", |
| "suffix": "" |
| }, |
| { |
| "first": "Morteza", |
| "middle": [], |
| "last": "Dehghani", |
| "suffix": "" |
| }, |
| { |
| "first": "Heng", |
| "middle": [], |
| "last": "Ji", |
| "suffix": "" |
| } |
| ], |
| "year": 2018, |
| "venue": "2018 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining (ASONAM)", |
| "volume": "", |
| "issue": "", |
| "pages": "552--559", |
| "other_ids": { |
| "DOI": [ |
| "10.1109/ASONAM.2018.8508244" |
| ] |
| }, |
| "num": null, |
| "urls": [], |
| "raw_text": "Ying Lin, Joe Hoover, Gwenyth Portillo-Wightman, Christina Park, Morteza Dehghani, and Heng Ji. 2018. Acquiring background knowledge to improve moral value prediction. In 2018 IEEE/ACM Interna- tional Conference on Advances in Social Networks Analysis and Mining (ASONAM), pages 552-559.", |
| "links": null |
| }, |
| "BIBREF16": { |
| "ref_id": "b16", |
| "title": "Syntactic annotations for the Google books NGram corpus", |
| "authors": [ |
| { |
| "first": "Yuri", |
| "middle": [], |
| "last": "Lin", |
| "suffix": "" |
| }, |
| { |
| "first": "Jean-Baptiste", |
| "middle": [], |
| "last": "Michel", |
| "suffix": "" |
| }, |
| { |
| "first": "Jon", |
| "middle": [], |
| "last": "Erez Aiden Lieberman", |
| "suffix": "" |
| }, |
| { |
| "first": "Will", |
| "middle": [], |
| "last": "Orwant", |
| "suffix": "" |
| }, |
| { |
| "first": "Slav", |
| "middle": [], |
| "last": "Brockman", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Petrov", |
| "suffix": "" |
| } |
| ], |
| "year": 2012, |
| "venue": "Proceedings of the ACL 2012 System Demonstrations", |
| "volume": "", |
| "issue": "", |
| "pages": "169--174", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Yuri Lin, Jean-Baptiste Michel, Erez Aiden Lieberman, Jon Orwant, Will Brockman, and Slav Petrov. 2012. Syntactic annotations for the Google books NGram corpus. In Proceedings of the ACL 2012 System Demonstrations, pages 169-174, Jeju Island, Korea. Association for Computational Linguistics.", |
| "links": null |
| }, |
| "BIBREF17": { |
| "ref_id": "b17", |
| "title": "Efficient estimation of word representations in vector space", |
| "authors": [ |
| { |
| "first": "Tomas", |
| "middle": [], |
| "last": "Mikolov", |
| "suffix": "" |
| }, |
| { |
| "first": "Kai", |
| "middle": [], |
| "last": "Chen", |
| "suffix": "" |
| }, |
| { |
| "first": "Greg", |
| "middle": [], |
| "last": "Corrado", |
| "suffix": "" |
| }, |
| { |
| "first": "Jeffrey", |
| "middle": [], |
| "last": "Dean", |
| "suffix": "" |
| } |
| ], |
| "year": 2013, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": { |
| "arXiv": [ |
| "arXiv:1301.3781" |
| ] |
| }, |
| "num": null, |
| "urls": [], |
| "raw_text": "Tomas Mikolov, Kai Chen, Greg Corrado, and Jef- frey Dean. 2013. Efficient estimation of word representations in vector space. arXiv preprint arXiv:1301.3781.", |
| "links": null |
| }, |
| "BIBREF18": { |
| "ref_id": "b18", |
| "title": "Understanding rating behavior based on moral foundations: The case of yelp reviews", |
| "authors": [ |
| { |
| "first": "Pegah", |
| "middle": [], |
| "last": "Nokhiz", |
| "suffix": "" |
| }, |
| { |
| "first": "Fengjun", |
| "middle": [], |
| "last": "Li", |
| "suffix": "" |
| } |
| ], |
| "year": 2017, |
| "venue": "2017 IEEE International Conference on Big Data (Big Data)", |
| "volume": "", |
| "issue": "", |
| "pages": "3938--3945", |
| "other_ids": { |
| "DOI": [ |
| "10.1109/BigData.2017.8258402" |
| ] |
| }, |
| "num": null, |
| "urls": [], |
| "raw_text": "Pegah Nokhiz and Fengjun Li. 2017. Understanding rating behavior based on moral foundations: The case of yelp reviews. In 2017 IEEE International Conference on Big Data (Big Data), pages 3938- 3945.", |
| "links": null |
| }, |
| "BIBREF19": { |
| "ref_id": "b19", |
| "title": "GloVe: Global vectors for word representation", |
| "authors": [ |
| { |
| "first": "Jeffrey", |
| "middle": [], |
| "last": "Pennington", |
| "suffix": "" |
| }, |
| { |
| "first": "Richard", |
| "middle": [], |
| "last": "Socher", |
| "suffix": "" |
| }, |
| { |
| "first": "Christopher", |
| "middle": [], |
| "last": "Manning", |
| "suffix": "" |
| } |
| ], |
| "year": 2014, |
| "venue": "Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP)", |
| "volume": "", |
| "issue": "", |
| "pages": "1532--1543", |
| "other_ids": { |
| "DOI": [ |
| "10.3115/v1/D14-1162" |
| ] |
| }, |
| "num": null, |
| "urls": [], |
| "raw_text": "Jeffrey Pennington, Richard Socher, and Christopher Manning. 2014. GloVe: Global vectors for word representation. In Proceedings of the 2014 Confer- ence on Empirical Methods in Natural Language Processing (EMNLP), pages 1532-1543, Doha, Qatar. Association for Computational Linguistics.", |
| "links": null |
| }, |
| "BIBREF20": { |
| "ref_id": "b20", |
| "title": "Enhancing the measurement of social effects by capturing morality", |
| "authors": [ |
| { |
| "first": "Rezvaneh", |
| "middle": [], |
| "last": "Rezapour", |
| "suffix": "" |
| }, |
| { |
| "first": "H", |
| "middle": [], |
| "last": "Saumil", |
| "suffix": "" |
| }, |
| { |
| "first": "Jana", |
| "middle": [], |
| "last": "Shah", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Diesner", |
| "suffix": "" |
| } |
| ], |
| "year": 2019, |
| "venue": "Proceedings of the Tenth Workshop on Computational Approaches to Subjectivity, Sentiment and Social Media Analysis", |
| "volume": "", |
| "issue": "", |
| "pages": "35--45", |
| "other_ids": { |
| "DOI": [ |
| "10.18653/v1/W19-1305" |
| ] |
| }, |
| "num": null, |
| "urls": [], |
| "raw_text": "Rezvaneh Rezapour, Saumil H. Shah, and Jana Diesner. 2019. Enhancing the measurement of social effects by capturing morality. In Proceedings of the Tenth Workshop on Computational Approaches to Subjec- tivity, Sentiment and Social Media Analysis, pages 35-45, Minneapolis, USA. Association for Compu- tational Linguistics.", |
| "links": null |
| }, |
| "BIBREF21": { |
| "ref_id": "b21", |
| "title": "Moral rhetoric in twitter: A case study of the u.s. federal shutdown of", |
| "authors": [ |
| { |
| "first": "Eyal", |
| "middle": [], |
| "last": "Sagi", |
| "suffix": "" |
| }, |
| { |
| "first": "Morteza", |
| "middle": [], |
| "last": "Dehghani", |
| "suffix": "" |
| } |
| ], |
| "year": 2013, |
| "venue": "Cognitive Science", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Eyal Sagi and Morteza Dehghani. 2014. Moral rhetoric in twitter: A case study of the u.s. federal shutdown of 2013. Cognitive Science, 36.", |
| "links": null |
| }, |
| "BIBREF22": { |
| "ref_id": "b22", |
| "title": "The morality machine: Tracking moral values in tweets", |
| "authors": [ |
| { |
| "first": "Livia", |
| "middle": [], |
| "last": "Teernstra", |
| "suffix": "" |
| }, |
| { |
| "first": "Peter", |
| "middle": [], |
| "last": "Van Der Putten", |
| "suffix": "" |
| } |
| ], |
| "year": 2016, |
| "venue": "Liesbeth Noordegraaf-Eelens, and Fons Verbeek", |
| "volume": "", |
| "issue": "", |
| "pages": "26--37", |
| "other_ids": { |
| "DOI": [ |
| "10.1007/978-3-319-46349-0_3" |
| ] |
| }, |
| "num": null, |
| "urls": [], |
| "raw_text": "Livia Teernstra, Peter van der Putten, Liesbeth Noordegraaf-Eelens, and Fons Verbeek. 2016. The morality machine: Tracking moral values in tweets. In Advances in Intelligent Data Analysis XV, pages 26-37, Cham. Springer International Publishing.", |
| "links": null |
| }, |
| "BIBREF23": { |
| "ref_id": "b23", |
| "title": "Red, white, and blue enough to be green: Effects of moral framing on climate change attitudes and conservation behaviors", |
| "authors": [ |
| { |
| "first": "Christopher", |
| "middle": [], |
| "last": "Wolsko", |
| "suffix": "" |
| }, |
| { |
| "first": "Hector", |
| "middle": [], |
| "last": "Ariceaga", |
| "suffix": "" |
| }, |
| { |
| "first": "Jesse", |
| "middle": [], |
| "last": "Seiden", |
| "suffix": "" |
| } |
| ], |
| "year": 2016, |
| "venue": "Journal of Experimental Social Psychology", |
| "volume": "65", |
| "issue": "", |
| "pages": "7--19", |
| "other_ids": { |
| "DOI": [ |
| "10.1016/j.jesp.2016.02.005" |
| ] |
| }, |
| "num": null, |
| "urls": [], |
| "raw_text": "Christopher Wolsko, Hector Ariceaga, and Jesse Sei- den. 2016. Red, white, and blue enough to be green: Effects of moral framing on climate change attitudes and conservation behaviors. Journal of Experimen- tal Social Psychology, 65:7-19.", |
| "links": null |
| }, |
| "BIBREF24": { |
| "ref_id": "b24", |
| "title": "Text-based inference of moral sentiment change", |
| "authors": [ |
| { |
| "first": "Jing", |
| "middle": [], |
| "last": "Yi Xie", |
| "suffix": "" |
| }, |
| { |
| "first": "Renato Ferreira Pinto", |
| "middle": [], |
| "last": "Junior", |
| "suffix": "" |
| }, |
| { |
| "first": "Graeme", |
| "middle": [], |
| "last": "Hirst", |
| "suffix": "" |
| }, |
| { |
| "first": "Yang", |
| "middle": [], |
| "last": "Xu", |
| "suffix": "" |
| } |
| ], |
| "year": 2019, |
| "venue": "Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP)", |
| "volume": "", |
| "issue": "", |
| "pages": "4654--4663", |
| "other_ids": { |
| "DOI": [ |
| "10.18653/v1/D19-1472" |
| ] |
| }, |
| "num": null, |
| "urls": [], |
| "raw_text": "Jing Yi Xie, Renato Ferreira Pinto Junior, Graeme Hirst, and Yang Xu. 2019. Text-based inference of moral sentiment change. In Proceedings of the 2019 Conference on Empirical Methods in Natu- ral Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), pages 4654-4663, Hong Kong, China. Association for Computational Linguistics.", |
| "links": null |
| } |
| }, |
| "ref_entries": { |
| "TABREF1": { |
| "html": null, |
| "text": "Number of concepts per moral trait.", |
| "type_str": "table", |
| "num": null, |
| "content": "<table/>" |
| }, |
| "TABREF3": { |
| "html": null, |
| "text": "Prediction accuracy for all moral traits, KGs and PR methods scores for each PPR method.", |
| "type_str": "table", |
| "num": null, |
| "content": "<table/>" |
| }, |
| "TABREF5": { |
| "html": null, |
| "text": "Confusion matrix of moral trend prediction on WordNet, for Uniform PPR", |
| "type_str": "table", |
| "num": null, |
| "content": "<table><tr><td>Predicted class</td></tr></table>" |
| }, |
| "TABREF6": { |
| "html": null, |
| "text": "Confusion matrix for the prediction of moral traits on DBpedia, for Type-Exclusivity PPR mance is still better than all the methods proposed byXie et al. (2019) when training the MFD lexicon entry embeddings on the COHA corpus 2 .", |
| "type_str": "table", |
| "num": null, |
| "content": "<table/>" |
| }, |
| "TABREF8": { |
| "html": null, |
| "text": "Accuracy for all moral foundations, KGs and PPR methods", |
| "type_str": "table", |
| "num": null, |
| "content": "<table><tr><td/><td>Authority /</td><td>Care</td><td>/</td><td>Fairness /</td><td>Loyalty /</td><td>Purity</td><td>/</td><td>General</td></tr><tr><td/><td>Subversion</td><td>Harm</td><td/><td>Cheating</td><td>Betrayal</td><td colspan=\"2\">Degradation</td><td>morality</td></tr><tr><td>Authority / Subversion</td><td>.60*</td><td>.09</td><td/><td>.05</td><td>.11</td><td>.07</td><td/><td>.07</td></tr><tr><td>Care / Harm</td><td>.05</td><td>.72*</td><td/><td>.04</td><td>.07</td><td>.05</td><td/><td>.06</td></tr><tr><td>Fairness / Cheating</td><td>.06</td><td>.06</td><td/><td>.59*</td><td>.09</td><td>.06</td><td/><td>.14</td></tr><tr><td>Loyalty / Betrayal</td><td>.13</td><td>.11</td><td/><td>.10</td><td>.59*</td><td>.03</td><td/><td>.03</td></tr><tr><td>Purity / Degradation</td><td>.07</td><td>.11</td><td/><td>.04</td><td>.00</td><td>.54*</td><td/><td>.24</td></tr><tr><td>General morality</td><td>.04</td><td>.05</td><td/><td>.05</td><td>.01</td><td>.15</td><td/><td>.71*</td></tr><tr><td/><td colspan=\"5\">(a) KG: WordNet; Method: Uniform PPR</td><td/><td/></tr><tr><td/><td>Authority /</td><td>Care</td><td>/</td><td>Fairness /</td><td>Loyalty /</td><td>Purity</td><td>/</td><td>General</td></tr><tr><td/><td>Subversion</td><td>Harm</td><td/><td>Cheating</td><td>Betrayal</td><td colspan=\"2\">Degradation</td><td>morality</td></tr><tr><td>Authority / Subversion</td><td>.49*</td><td>.08</td><td/><td>.13</td><td>.10</td><td>.08</td><td/><td>.11</td></tr><tr><td>Care / Harm</td><td>.10</td><td>.55*</td><td/><td>.11</td><td>.10</td><td>.10</td><td/><td>.04</td></tr><tr><td>Fairness / Cheating</td><td>.08</td><td>.03</td><td/><td>.53*</td><td>.17</td><td>.03</td><td/><td>.15</td></tr><tr><td>Loyalty / Betrayal</td><td>.14</td><td>.14</td><td/><td>.16</td><td>.37*</td><td>.04</td><td/><td>.14</td></tr><tr><td>Purity / Degradation</td><td>.07</td><td>.02</td><td/><td>.05</td><td>.05</td><td>.63*</td><td/><td>.19</td></tr><tr><td>General morality</td><td>.00</td><td>.00</td><td/><td>.20</td><td>.00</td><td>.08</td><td/><td>.72*</td></tr><tr><td/><td colspan=\"5\">(b) KG: DBpedia; Method: Type-Exclusivity PPR</td><td/><td/></tr></table>" |
| }, |
| "TABREF9": { |
| "html": null, |
| "text": "Confusion matrices for predicting the moral foundation", |
| "type_str": "table", |
| "num": null, |
| "content": "<table/>" |
| }, |
| "TABREF11": { |
| "html": null, |
| "text": "Accuracy for moral polarity, KGs and PPR methods", |
| "type_str": "table", |
| "num": null, |
| "content": "<table><tr><td>F-virtue</td><td>F-vice</td><td>A-virtue</td><td>A-vice</td></tr><tr><td>equitable, just</td><td>subjective</td><td>servile</td><td>resistance</td></tr><tr><td>nonpartisan, nonpartizan</td><td>unprincipled</td><td>position, situation</td><td>unorthodox</td></tr><tr><td>democratic</td><td>partiality, partisanship</td><td>attitude, mental attitude</td><td>jurisprudence, law</td></tr><tr><td>mutual, reciprocal</td><td>disposition, inclination</td><td>follower</td><td>intractability, intractableness</td></tr><tr><td>disposition, inclination</td><td>act upon, influence</td><td>admirer, champion</td><td>dissent, resist</td></tr><tr><td>broad-minded</td><td>omission</td><td>reputable</td><td>loyalty, trueness</td></tr><tr><td>ism, philosophical system</td><td>corrupt</td><td>honorable, honourable</td><td>bad hat, mischief-maker</td></tr><tr><td>conformance, conformity</td><td>intolerant</td><td>tenderness, warmness</td><td>uncontrolled</td></tr><tr><td>true, truthful</td><td>ideology, political orientation</td><td>pious</td><td>provocative</td></tr><tr><td>just</td><td>advantage, vantage</td><td>courteous</td><td>disloyal, unpatriotic</td></tr><tr><td colspan=\"4\">(a) WN: For space reasons, we only show the two first words of each synset.</td></tr><tr><td>F-virtue</td><td>F-vice</td><td>A-virtue</td><td>A-vice</td></tr><tr><td>justify</td><td>judgment</td><td>ranke</td><td>heresy</td></tr><tr><td>bias</td><td>nonstandard</td><td>slang</td><td>ick</td></tr><tr><td>fair minded</td><td>inequality</td><td>person</td><td>legal</td></tr><tr><td>fair mindedly</td><td>judge</td><td>us</td><td>disagreement</td></tr><tr><td>prejudice</td><td>raptophilia</td><td>computing</td><td>outlaw</td></tr><tr><td>fair mindedness</td><td>out of proportion</td><td>detraditionalize</td><td>defier</td></tr><tr><td>just</td><td>fair</td><td>historical</td><td>law</td></tr><tr><td>philosophy</td><td>nonsegregational</td><td>maternal</td><td>person</td></tr><tr><td>right</td><td>separate</td><td>honourable</td><td>obedience</td></tr><tr><td>nonjustificational</td><td>equal</td><td>honorarium</td><td>us</td></tr><tr><td/><td colspan=\"2\">(b) CN</td><td/></tr><tr><td>F-virtue</td><td>F-vice</td><td>A-virtue</td><td>A-vice</td></tr><tr><td>National debt of the United</td><td>Frank Stanford</td><td>Obedience to Authority: An</td><td>Jerome Brailey</td></tr><tr><td>States</td><td>Persecution of Ahmadis</td><td>Experimental View</td><td>Petty treason</td></tr><tr><td>Freedom of Religion</td><td>Princelings</td><td>What Comes After Goodbye</td><td>Siege of Lier (1582)</td></tr><tr><td>Work motivation</td><td>Racial wage gap in the</td><td>Suprematism</td><td>Private Lies (book)</td></tr><tr><td>Coretta Scott King</td><td>United States</td><td>Standings</td><td>Descent</td></tr><tr><td>Al-Baqara 256</td><td>Blacklisting</td><td>Robert Holden (author)</td><td>Version Control Example</td></tr><tr><td>A Critique of Pure Tolerance</td><td>Shunning</td><td>Filial piety</td><td>Sedation</td></tr><tr><td>Zechariah Chafee</td><td>United States</td><td>Blondes (John Stewart</td><td>Civil Rights Act of 1968</td></tr><tr><td>Horizontal inequality</td><td>Mick Moore (political</td><td>album)</td><td>Universum (band)</td></tr><tr><td>Life estate</td><td>economist)</td><td>Emil Hassler</td><td>The Politics of Religious</td></tr><tr><td>Human rights in the Middle</td><td>Barbara Risman</td><td>United States</td><td>Apostasy</td></tr><tr><td>East</td><td>Lahore Grammar School</td><td>Legal Legitimacy</td><td/></tr><tr><td/><td>Multan</td><td/><td/></tr><tr><td/><td colspan=\"2\">(c) DBpedia</td><td/></tr></table>" |
| }, |
| "TABREF12": { |
| "html": null, |
| "text": "Top-10 non-seed concepts for every KB", |
| "type_str": "table", |
| "num": null, |
| "content": "<table/>" |
| } |
| } |
| } |
| } |
