phenomenon · claim-type · cross-domain-transfer · mechanism · claim-relation · replication-status · limitation — grounded in the full text| phenomenon | figurative language as a network of topics— “This process of meaning transference through metaphors, which takes place in relation to a previous unwitting layer of metaphors, I argue, has the structure of a network, a network that forms and reforms out of the various interactions between the metaphoric components.” | |
| claim-type | method— “This article suggests a computational method for the analysis of figurative language as a network, using methods from natural language processing (NLP), particularly topic classification and network analysis.” | |
| method — evidence | — “Although my focus is on one case study, namely Timon of Athens, and hence a manual classification would have been enough to test this method's potential for textual analysis, I have trained a topic classifier both to demonstrate the computational automatization of topic classification and to provide a basis for future research on a large… | |
| transfer | from natural language processing: the computational apparatus of natural language processing - tokenisation, distributional similarity, network construction - applied to figurative language in a corpus of liturgical poetry— “This article suggests a computational method for the analysis of figurative language as a network, using methods from natural language processing (NLP), particularly topic classification and network analysis.” | transfers[0] |
| mechanism | THE DIFFICULTY IS ATTRIBUTED TO THE STRUCTURE OF THE OBJECT, NOT TO THE METHOD: analysing figurative RATHER THAN LITERAL language POSES CERTAIN DIFFICULTIES DUE TO THE VERY STRUCTURE OF FIGURATIVE EXPRESSIONS - so the obstacle is located in what is being measured before any tool is named— “The analysis of figurative language, rather than literal language, poses certain difficulties due to the very structure of figurative expressions. The suggested method in this articles offers a way to overcome these difficulties.” | mechanisms[0] |
| mechanism | A PROHIBITION IS SHOWN VIOLATING ITSELF IN ITS OWN WORDING: the Areopagite judges CONDEMN the use of figurative language AS DECEPTIVE, and THEY THEMSELVES USE A METAPHOR OF A CARPENTER IN THEIR CONDEMNATION. The rule against a class of language is stated in that class of language— “Ironically, although the Areopagite judges condemn the use of figurative language as deceptive, they themselves use a metaphor of a carpenter in their condemnation” | mechanisms[1] |
| mechanism | AND THE SHARPER POINT IS THAT THE HOMONYMY DOES THE WORK, NOT THE METAPHOR: the influence is NOT ONLY evident through the carpenter - RATHER IT IS THE MORE SUBTLE HOMONYM SENSES OF THE WORDS UPRIGHT AND RULE THAT TRULY REVEAL THE PERVASIVE POWER OF THE FIGURATIVE. Two words carrying two senses each, unnoticed by the speakers, are what the argument turns on— “The influence of figurative language is not only evident here through the metaphor of the carpenter, but rather it is the more subtle homonym senses of the words upright and rule that truly reveal the pervasive power of the figurative.” | mechanisms[2] |
| mechanism | A CLASS OF METAPHOR IS DEFINED BY WHAT WOULD BE IMPOSSIBLE WITHOUT IT: the word is catalogued as part of the ORIENTATIONAL class of metaphors, which RELY ON OUR MOST BASIC SENSE OF SPACE AND PHYSICAL CONCEPTS, so much so that WITHOUT WHICH WE COULD NOT FUNCTION IN THE WORLD - COULD NOT REASON OR COMMUNICATE. The category's members are identified by the cost of their absence— “In Metaphors We Live By, Lakoff and Johnson catalogue the word upright as part of the orientational class of metaphors, which rely on our most basic sense of space and physical concepts, so much so that “without which we could not function in the world – could not reason or communicate”” | mechanisms[3] |
| mechanism | A METHOD'S OUTPUT AND THE CONCEPT IT IS MEANT TO MEASURE ARE REPORTED NOT TO CORRESPOND: after comparing a computational reading with a thematic one, the conclusion quoted is WE CANNOT FIND A CLEAR OR CONSTANT RELATIONSHIP BETWEEN AN LDA-TOPIC AND A THEME. The instrument produces something, and it is not the thing the word names— “he concluded that “we cannot find a clear or constant relationship between an LDA-topic and a theme out of this comparison” [Navarro-Colorado 2018, 7].” | mechanisms[4] |
| mechanism | THE METHOD IS SAID TO WORK ON ONE TEXT TYPE AND FAIL ON ANOTHER, AND THE REASON GIVEN IS CONTEXT-EXCLUSIVITY OF VOCABULARY: topic modelling applied to poetry DOES NOT EXTRACT CLEAR AND DEFINED THEMES AS IT DOES WHEN APPLIED TO SCIENTIFIC TEXTS, because IN SCIENTIFIC TEXTS THERE ARE WORDS THAT ARE USED EXCLUSIVELY IN VERY SPECIFIC CONTEXTS - and THE LINGUISTIC USES IN POETRY WORK IN THE EXACT OPPOSITE WAY. The property that makes the method work is a property of the vocabulary, not of the algorithm— “LDA topic modeling applied to poetry does not extract clear and defined themes as it does when applied to scientific texts. […] In scientific texts there are words that are used exclusively in very specific contexts, related to specific topics. […] The linguistic uses in poetry work in the exact opposite way.” | mechanisms[5] |
| mechanism | THE NAMING STEP IS IDENTIFIED AS THE MANUAL AND SUBJECTIVE ONE: in topic modelling, DISCERNING THE TITLES of the groups of words REMAINS A MANUAL TASK THAT REQUIRES ACTIVE AND OFTEN SUBJECTIVE INTERPRETATION. The algorithm delivers groups; a person supplies the words that say what the groups are— “In topic modeling, however, discerning the titles of the groups of words remains a manual task that requires active and often subjective interpretation.” | mechanisms[6] |
| mechanism | A NAMED DEFECT OF AUTOMATIC GROUPING IS CITED, AND IT IS A DISAGREEMENT BETWEEN MACHINE AND HUMAN CATEGORIES: a list of words treated as one topic MAY INCLUDE WORDS THAT HUMAN INTERPRETERS RECOGNIZE AS BELONGING TO DISTINCT TOPICS - a problem with its own name, WORD INTRUSION. The defect is defined by the mismatch, and the mismatch is what the name records— “A list of words considered as a topic by a topic modeling may include words that human interpreters recognize as belonging to distinct topics, a problem described by Chang et al. as “word intrusion”” | mechanisms[7] |
| mechanism | THE RIGHT RELIABILITY MEASURE IS NAMED, ITS PURPOSE STATED, AND IT IS EXPLICITLY NOT DONE, WITH THE REASON: given the SAME set of topics and instances, TWO INTERPRETERS MIGHT STILL CHOOSE DIFFERENT TOPICS for a word - INTER-ANNOTATORS AGREEMENT is the method by which that agreement can be measured, PROVIDING A STATUS OF CONSENSUS, and the procedure REQUIRES CERTAIN RESOURCES AND HENCE WILL BE IMPLEMENTED later. The measure that would qualify the result is named and deferred in the same breath— “Another subjective aspect pertains to the classifying process itself. Given the same set of topics and figurative instances, two interpreters might still choose different topics for a word in certain cases. The inter-annotators agreement (IAA) is a method by which the agreement between interpreters can be measured, providing a status of… | mechanisms[8] |
| mechanism | THE IMPROVEMENTS PROPOSED ARE PROPERTIES OF THE SETUP RATHER THAN OF THE ANALYSIS: the recommendations begin with REDUCING THE NUMBER OF TOPICS AND ENLARGING THE DATASET, which will allow the algorithm to have MORE DATA AND LESS AMBIGUITY REGARDING TOPICS - so the ambiguity is treated as a quantity the design can lower rather than a feature of the object— “Several recommendations can be made based on training the classifier to reach better results, starting with reducing the number of topics and enlarging the dataset, which will allow the algorithm to have more data and less ambiguity regarding topics.” | mechanisms[9] |
| builds-on | Moretti's notion of "operationalization" (Moretti 2013), after Bridgman's The Logic of Modern Physics— “This process is described by Moretti as “operationalization”, applying to digital humanities the work of Bridgman in the philosophy of science, The Logic of Modern Physics.” | claim_relations[0] |
| builds-on | Lakoff and Johnson (2003), Metaphors We Live By — metaphorical entailment— “That metaphors have the structure of a network is arguably most evident in Lakoff and Johnson's notion of metaphorical entailment, according to which instances of a structural metaphor such as “time is money” (e.g. “time is a resource”, “time is a valuable commodity”), link between “all the instances of a single metaphorical structuring… | claim_relations[1] |
| contradicts | Rhody (2012), topic modeling of figurative language— “Although Rhody suggests a way to benefit from topic modeling in the analysis of figurative texts, this merely bypasses the difficulties rather than solving them.” | claim_relations[2] |
| contradicts | Finkelstein (2020), "Amicitia and beneficia in Timon of Athens"— “In the network of Timon of Athens, the topic of “politics” is relatively a minor one, ranking between #18-20 among the various centralities (see Table 2); hence, we may question the stress that Finklestein puts on the political aspects of the play.” | claim_relations[3] |
| sense | a topic title - SUPPLIED BY A READER, AND DEFENDED AS SUCH: the paper states that titling the groups of words is a manual task requiring active and often subjective interpretation, and then argues that this subjectivity does not undermine the method but shows that various classifications are possible depending on the researcher's orientation - a position, not an oversight— “Although determining the number of topics and their titles is partly subjective, the subjectivity of this process does not necessarily undermine the viability of topic classification. Rather, it shows that various classifications are possible depending on the researcher's orientation” | senses[0] |
| sense | tenor and vehicle - GLOSSED IN ONE PARENTHESIS: the underlying subject and the transported meaning, taken from I. A. Richards; the paper states the connection between them is not a necessary one but rests on substitution, which is why the network it builds is undirected— “The connection between the tenor and the vehicle of a metaphor (i.e., the underlying subject and the transported meaning) is not a necessary one, but rather based on substitution and a relative freedom from any constraints that a given context may place on a word.” | senses[1] |
| sense | word intrusion - THE NAMED FAILURE MODE, ATTRIBUTED: a list of words the model calls one topic may contain words that human interpreters place in distinct topics, which the paper attributes to Chang et al. rather than presenting as its own discovery— “In topic modeling, however, discerning the titles of the groups of words remains a manual task that requires active and often subjective interpretation. A list of words considered as a topic by a topic modeling may include words that human interpreters recognize as belonging to distinct topics, a problem described by Chang et al. as “word… | senses[2] |
| control | randomly splitting the dataset into a 66% training set and a 33% test set — held— “In the process of training the algorithm, the dataset is randomly split into two sets, a training dataset (66%) and a test dataset (33%), so that the algorithm can compare between its own predictions and the original ones made by the creator of the dataset, and accordingly the accuracy is calculated.” — rules out that the classifier's apparent accuracy comes from memorising the data it was trained on— “In the process of training the algorithm, the dataset is randomly split into two sets, a training dataset (66%) and a test dataset (33%), so that the algorithm can compare between its own predictions and the original ones made by the creator of the dataset, and accordingly the accuracy is calculated.” | controls[0] |
| differs in | Rhody (2012), topic modeling of figurative language — BYPASSING AGAINST SOLVING. 'Although Rhody suggests a way to benefit from topic modeling in the analysis of figurative texts, this merely bypasses the difficulties rather than solving them.' The method is granted as usable; what is contested is whether it addresses the problem.— “Although Rhody suggests a way to benefit from topic modeling in the analysis of figurative texts, this merely bypasses the difficulties rather than solving them.”— “this merely bypasses the difficulties rather than solving them” | claim_relations[2] |
| differs in | tenor —distinct-from→ vehicle — which side of the metaphor each names: the underlying subject, or the transported meaning— “The connection between the tenor and the vehicle of a metaphor (i.e., the underlying subject and the transported meaning) is not a necessary one, but rather based on substitution and a relative freedom from any constraints that a given context may place on a word.”— “the underlying subject and the transported meaning” | concept_edges[0] |
| outcome | control — held— “In the process of training the algorithm, the dataset is randomly split into two sets, a training dataset (66%) and a test dataset (33%), so that the algorithm can compare between its own predictions and the original ones made by the creator of the dataset, and accordingly the accuracy is calculated.” | controls[0] |
| for which use | limitation — any topic produced by the modelling step here: the authors note that discerning the titles is a manual task requiring subjective interpretation, and that a topic's word list may include words human interpreters recognise as belonging to distinct topics, a problem named word intrusion— “In topic modeling, however, discerning the titles of the groups of words remains a manual task that requires active and often subjective interpretation. A list of words considered as a topic by a topic modeling may include words that human interpreters recognize as belonging to distinct topics, a problem described by Chang et al. as “word… | limitations[7] |
| evidence status note | mechanism — The span contains the author's typo "in this articles" for "in this article". Not extraction damage - it is in the published DHQ text - and quoted verbatim.— “The analysis of figurative language, rather than literal language, poses certain difficulties due to the very structure of figurative expressions. The suggested method in this articles offers a way to overcome these difficulties.” | mechanisms[0] |
| evidence status note | mechanism — EXTENDED to the full stop, and the span contains the author's typo "Tthis procedure" with a doubled T. Not extraction damage; it is in the published text. The extension matters because "will be implemented in a future study" is what dates the deferral - the reliability measure is named, its purpose stated, and its absence scheduled, all in one passage.— “Another subjective aspect pertains to the classifying process itself. Given the same set of topics and figurative instances, two interpreters might still choose different topics for a word in certain cases. The inter-annotators agreement (IAA) is a method by which the agreement between interpreters can be measured, providing a status of… | mechanisms[8] |
| replication | untested | |
| limitation | Figurative instances still have to be extracted from the plays by hand; no automatic procedure isolates them.— “For the time being, it seems there is no escape from extracting the figurative instances manually.” | limitations[0] |
| limitation | The trained topic classifier reaches only 72.94% accuracy and is presented as needing further fine-tuning.— “The topic classifier currently achieves a promising result of 72.94% accuracy (the accuracy is the percentage of correct predictions of the classifier), which can be further improved for future work by fine-tuning the machine learning process.” | limitations[1] |
| limitation | The number of topics and their titles are partly subjective researcher decisions.— “Although determining the number of topics and their titles is partly subjective, the subjectivity of this process does not necessarily undermine the viability of topic classification.” | limitations[2] |
| limitation | Inter-annotator agreement on the manual classification was not measured; it is deferred to a future study.— “Tthis procedure requires certain resources and hence will be implemented in a future study.” | limitations[3] |
| limitation | The automatic classification is described as still needing improvement (fewer topics, larger dataset, other language models).— “Though the automatic classification of topics in figurative instances has shown promising results, there are still improvements that can be made.” | limitations[4] |
| limitation | Extending the network to a directed one (tenor to vehicle) is blocked by cases where the tenor is only implied.— “This, however, introduces a variety of challenges, such as figurative cases in which the tenor is absent textually and only implied.” | limitations[5] |
| limitation | Low centrality in the network does not establish that a topic is unimportant, so the method cannot by itself dismiss an interpretation.— “That a certain topic appears a low number of times does not necessarily mean it is not meaningful, as it could appear in crucial parts of a text.” | limitations[6] |
| limitation | A modelled topic may group words that human readers see as belonging to different topics. | limitations[7] |
finding · quantity · variable · definition · key-term — grounded in the full text| finding | Figurative language is harder to analyse than literal language because of the very structure of figurative expressions. | content_tags.findings[0] |
| evidence status note | finding — QUOTE RECOVERED FROM THE PAPER'S OWN CHARACTERS. The recorded quote was LETTERS-ONLY IDENTICAL to this stretch of the staged text but not an exact substring of it - spaces, hyphens, ligature glyphs or mathematical operators differed, so it could never anchor. This is the file's own text between the same two letters: not produced by a rule about the damage, copied out of the file. Round-trip checked (the recovered span is letters-only equal to the original) and length-checked. The text as originally recorded is kept in evidence as extracted. | content_tags.findings[0] |
| finding | The method combines two techniques rather than one: topic classification to label the figures, network analysis to relate them. | content_tags.findings[1] |
| evidence status note | finding — QUOTE RECOVERED FROM THE PAPER'S OWN CHARACTERS. The recorded quote was LETTERS-ONLY IDENTICAL to this stretch of the staged text but not an exact substring of it - spaces, hyphens, ligature glyphs or mathematical operators differed, so it could never anchor. This is the file's own text between the same two letters: not produced by a rule about the damage, copied out of the file. Round-trip checked (the recovered span is letters-only equal to the original) and length-checked. The text as originally recorded is kept in evidence as extracted. | content_tags.findings[1] |
| finding | Figurative analysis is not confined to literary criticism; it matters wherever the object of study is text. | content_tags.findings[2] |
| evidence status note | finding — QUOTE RECOVERED FROM THE PAPER'S OWN CHARACTERS. The recorded quote was LETTERS-ONLY IDENTICAL to this stretch of the staged text but not an exact substring of it - spaces, hyphens, ligature glyphs or mathematical operators differed, so it could never anchor. This is the file's own text between the same two letters: not produced by a rule about the damage, copied out of the file. Round-trip checked (the recovered span is letters-only equal to the original) and length-checked. The text as originally recorded is kept in evidence as extracted. | content_tags.findings[2] |
| finding | The method is judged by whether it can support a close reading, not only by whether it runs. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.findings[3] |
| evidence status note | finding — QUOTE RECOVERED FROM THE PAPER'S OWN CHARACTERS. The recorded quote was LETTERS-ONLY IDENTICAL to this stretch of the staged text but not an exact substring of it - spaces, hyphens, ligature glyphs or mathematical operators differed, so it could never anchor. This is the file's own text between the same two letters: not produced by a rule about the damage, copied out of the file. Round-trip checked (the recovered span is letters-only equal to the original) and length-checked. The text as originally recorded is kept in evidence as extracted. | content_tags.findings[3] |
| key-terms | figurative language, topic classification, network analysis, close reading, metaphor, literary criticism |
| concept | tenor Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[5] |
| concept | vehicle Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[6] |
| concept | word intrusion Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[7] |
| concept | topic modeling Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[8] |
When we have changed our mind. Some entries in the table begin with a word in capitals and the reason for it. EVIDENCE CORRECTED — the finding still stands, but we had quoted the wrong sentence from the paper, and the quotation has been replaced with the right one. RETRACTED — the entry itself was wrong; it stays on the page, marked, with the reason, rather than being deleted. WITHDRAWN AND REPLACED — the entry was wrong, and a corrected version of it appears in the entries below. Each mark carries its own reason, in brackets, because the reasons differ from one entry to the next. An entry we have WITHDRAWN is removed from this page and kept in the record, with its reason. A line reading Audit and a date is a later re-reading of the entry against its whole sentence: DOUBTFUL means it was kept, and the reason says what is loose about it. Quotation marks inside the tagger's sentence. Where a tag is the tagger's own sentence, a phrase in quotation marks inside it is borrowed from the paper — the tagger's wording around it is not. The paper's own sentence is the evidence quote for that row; on this page it is withheld with the article, and the article is at its publisher. Nothing inside quotation marks anywhere on this page is ours: if it is quoted, it is the paper's. Whose words a concept carries. A concept entry records a term the paper itself uses, so the term, the sense given for it, and its evidence are the paper's own words, quoted where the paper fixes them. This is the opposite of a sense entry in the tag table, which is the tagger's sentence about how the paper uses a term. Where a concept entry could not be grounded in the paper's words, the entry says so on its own row.
Tag layer © the Meta-Tagging Project, CC BY 4.0. The paper text is © its authors/publisher
and is linked, not reproduced. Excerpts shown are short quotations for identification and verification.
· back to index
· concepts · authors
· עברית
© 2026 Shir Sivroni · ORCID 0009-0000-2597-2824