Individual Words Research Articles

Emotional context and predictability in naturalistic reading aloud.

[Correction Notice: An Erratum for this article was reported in Vol 24(3) of Emotion (see record 2024-72515-001). In the article (https://doi.org/10.1037/emo0001298), Table 1 as originally reported contained an error with respect to participant demographics. Specifically, participants who had selected multiple categories for race/ethnicity were mistakenly assigned to only the first alphabetical category selected. Updates have been made in the Race/ethnicity section of Table 1, to change the heading "American Indian or Alaska Native" to "Multiple selected," and to the relevant statistics under that heading as well as under the "Hispanic, Latino/a/x, or Spanish origin" and "Asian" headings. No inferential statistics are impacted by this correction, nor does it affect the results or conclusions of the article. All versions of this article have been corrected.] A robust experimental literature has found that word frequency and lexical valence contribute to visual word processing at the level of the individual word. Extensions of this literature to simplified sentences have essentially corroborated single-word findings, albeit with important influences of the unfolding discourse context, which may strengthen or attenuate single-word effects. This study sought to extend current knowledge one step further, beyond stand-alone sentences or sentence pairs, by investigating how word frequency and lexical valence, along with their interactions, influence oral reading performance for multisentence stimuli in a naturalistic context. Lexical features were averaged over short passages of text, which were presented to participants on-screen simultaneously, and performance was assessed as reading speed, in words per second. Overall, we find that the same patterns emerge for multisentence oral reading as in the prior literature: strong frequency effects that benefit higher-frequency content, a positivity bias that increases reading speed for more positive content, and an important interaction that disfavors relatively more negative (less positive), high-frequency content. We discuss these findings in light of possible interpretations based on associative connectivity in the mental lexicon, as well as oculomotor dynamics during naturalistic reading. Our data suggest that reading speed of multisentence texts is a viable alternative, and one that offers enhanced ecological validity, for investigations of visual word processing dynamics. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

ПРИЕМ ПОВТОРА В АЛТАЙСКИХ ГЕРОИЧЕСКИХ СКАЗАНИЯХ

Altai heroic tales are unique monuments of the people’s epic heritage. They act as a cultural code that preserved the originality of the ethnic worldview, traditions of the people, as well as the peculiarities of their poetic language. At the same time, the Altai epic has much in common and peculiar in comparison with the epics of other Turkic-speaking peoples. The study of the epic style originality is special interest for folkloristics. The relevance of this study is determined by the insufficient study of style features of the Altai people’s heroic tales. The scholarly novelty of the work consists in the fact that it is the first to study the types of repetition as one of the stylistic phenomena of the Altai epic. The purpose of this research is to study repetition as a stylistic phenomenon of folk poetics. The sources for the study were the heroic tales published in the series “Altai Bogatyrs”. An important criterion for selecting sources for the study was the observation of this phenomenon in authentic texts, where the features of the narrator’s living language are preserved to a greater extent. The study is based on the analysis of Altai heroic tales from different narrators and other folklore sources. The descriptive method, methods of comparison and generalization were also used in the work. As a result of the study, the types of repetitions found in the Altai epic were identified and described; their main functions were defined. It is established that repetition in folklore text is an important and multidimensional phenomenon. In heroic tales it is manifested at different levels. Repetitions of individual words and word combinations, phrases and whole episodes are distinguished. According to the frequency of use, there are twofold, threefold and more repetitions. Repetitions occur within a line, at the beginning, end or at the junction of poetic lines, forming such stylistic figures as anaphora, epiphora and epanaphora. Each type of repetition fulfills a certain function in the text. The study has shown that the functioning of repetitions in heroic tales differs in the activity of their use by narrators.

Anomalous diffusion analysis of semantic evolution in major Indo-European languages.

How do words change their meaning? Although semantic evolution is driven by a variety of distinct factors, including linguistic, societal, and technological ones, we find that there is one law that holds universally across five major Indo-European languages: that semantic evolution is subdiffusive. Using an automated pipeline of diachronic distributional semantic embedding that controls for underlying symmetries, we show that words follow stochastic trajectories in meaning space with an anomalous diffusion exponent α = 0.45 ± 0.05 across languages, in contrast with diffusing particles that follow α = 1. Randomization methods indicate that preserving temporal correlations in semantic change directions is necessary to recover strongly subdiffusive behavior; however, correlations in change sizes play an important role too. We furthermore show that strong subdiffusion is a robust phenomenon under a wide variety of choices in data analysis and interpretation, such as the choice of fitting an ensemble average of displacements or averaging best-fit exponents of individual word trajectories.

Short Sentences of Responsive Replies from Dialogues of ‘Hamlet’ in Russian Translations

The article explores dialogic units with responsive utterances containing brief narrative structures consisting of representing or substituting words, such as we do, it will, they are not, etc. The linguistic material is sourced from the text of Shakespeare’s play ‘Hamlet’ and its translations into Russian by N. Polev, A. Kroneberg in the 19th century, B. Pasternak, M. Lozinsky, A. Radlova in the 20th century, V. Ananyin, I. Peshkov in the 21st century. The lexical-grammatical and stylistic characteristics of concise English structures are discussed, along with the translators’ approaches to their interpretation. It is revealed that in the translation texts, representative sentences are reflected by repeating individual words of the initiating utterance (sledi [follow] — slezhu [I follow]), idiomatic expressions (chego net, togo net [what is not, that is not]), sentence-words (da, net, vernо [yes, no, correct]), combinations like (sovershenno vernо [absolutely correct], tochno tak [just like that]), imperative markers (ne somnevaytes’ [do not doubt]). The authors of Russian translations maintain the function of utterances with representative sentences in Shakespeare’s text: in terms of character interaction — as an informative or phatic speech element ensuring coherence in dialogues, in terms of audience impact — as a tool helping viewers better navigate the stage events.

How Toddlers Use Core and Fringe Vocabulary: What's in an Utterance?

Selecting vocabulary for preliterate individuals who use augmentative and alternative communication presents multiple challenges, as the number of symbols provided must be balanced with cognitive, motoric, and other needs. Prioritizing certain types of vocabulary thus becomes a necessity. For example, prioritizing core vocabulary-that is, words that are commonly used across a group of people and contexts-is a common practice that attempts to address some of these issues. However, most core vocabulary research to date has narrowly focused on individual word counts, ignoring other critical aspects of language development such as how vocabulary aligns with typical development and how children use core and fringe vocabulary within their utterances. Descriptive and inferential statistics were used to analyze 112 transcripts to describe how typically developing toddlers (aged 2.5 years) use core and fringe vocabulary within their utterances, in reference to a range of commonly used core vocabulary lists. Results indicated that the proportion of the toddlers' utterances that consisted of only core, only fringe, or core + fringe vocabulary varied dramatically depending on the size of the core vocabulary list used, with smaller core lists yielding few "core-only" utterances. Furthermore, utterances containing both core and fringe vocabulary were both grammatically and semantically superior to utterances containing only core or only fringe vocabulary, as evidenced by measures such as mean length of utterance and total number of words. Thus, relying on word frequency counts is an insufficient basis for selecting vocabulary for aided preliterate communicators.

How Much Can Machines Learn Finance from Chinese Text Data?

How much can we learn finance directly from text data? This paper presents a new framework for learning textual data based on the factor augmentation model and sparsity regularization, called the factor-augmented regularized model for prediction (FarmPredict), to let machines learn financial returns directly from news. FarmPredict allows the model itself to extract information directly from articles without predefined information, such as dictionaries or pretrained models as in most studies. Using unsupervised learned factors to augment the predictors would benefit our method with a “double-robust” feature: that the machine would learn to balance between individual words or text factors/topics. It also avoids the information loss of factor regression in dimensionality reduction. We apply our model to the Chinese stock market with a large proportion of retail investors by using Chinese news data to predict financial returns. We show that positive sentiments scored by our FarmPredict approach from news generate on average 83 basic points (bps) stock daily excess returns, and negative news has an adverse impact of 26 bps on the days of news announcements, where both effects can last for a few days. This asymmetric effect aligns well with the short-sale constraints in the Chinese equity market. The result shows that the machine-learned prediction does provide sizeable predictive power with an annualized return of 54% at most with a simple investment strategy. Compared with other statistical and machine learning methods, FarmPredict significantly outperforms them on model prediction and portfolio performance. Our study demonstrates the far-reaching potential of using machines to learn text data. This paper was accepted by Kay Giesecke, finance. Funding: This study was supported by the National Natural Science Foundation of China [Grants 71991471, 71991470, and 72204049], the National Key Research and Development Program [Grant 2020YFA0608604], the Shanghai Pujiang Scholar Project [Grant 21PJC010], the Shanghai Science Project [Grant 23692119300], and the China Postdoctoral Science Project [Grants 2019M650076 and 2020T130107]. Supplemental Material: The online appendix and data files are available at https://doi.org/10.1287/mnsc.2022.01468 .

Part-Of-Speech Tagging for Balochi Language: A Data driven application of Conditional Random Fields

Parts-of-Speech (POS) tagging involves the assignment of the correct part of speech or lexical category to individual words within a sentence in a natural language. This procedure holds significant in the field of Natural Language Processing (NLP) and find utility across a variety of NLP applications. Commonly, it constitutes the initial phase of natural language processing. Subsequent stages may encompass additional tasks such as chunking, parsing and more. Balochi stands as the predominant language in Balochistan,, ranking as the fourth most prevalent language in Pakistan. The field of natural language processing for Balochi is still in its nascent stages. In this research, we introduce an algorithm for Balochi part-of-speech tagging, leveraging machine learning techniques. The core of our approach relies on a Conditional Random Field model as the machine learning component. Careful consideration is given to selecting appropriate features for the CRF, taking into account the linguistic characteristics of Balochi. Balochi is currently considered a resource poor language, and thus, the available manually tagged data consists of only approximately 1500 sentences. The tagset used in this study created for research purpose, consisting of 16 different tags. The learning process incorporates tagged data. The algorithm demonstrates a high accuracy rate of 86.78% when applied to Balochi texts. The training corpus comprises 40000 words, while the test corpus contains 10000 words.

YouTube Comment Sentiment Classification System

With more than 2 billion viewers per month, YouTube is the most widely used video-sharing website worldwide. On this website, users can watch, upload, and share videos covering a wide range of subjects. YouTube comments include facts, opinions, and responses to videos in addition to starting discussions. The number of YouTube comments makes it difficult to manually analyze them all. The study of reading, understanding, and creating text in human languages encompasses a broad range of methods and techniques under the umbrella of natural language processing or NLP. The primary goal of the research is to find and analyze YouTube comments, which, when used with natural language processing algorithms, might be beneficial for the channels' continued development. One of the NLP methods used for this research was tokenization, which is used to break down text into individual words or tokens. Stemming and lemmatization are used to reduce the root words and normalize the variation. Categorization is performed by identifying the named entities, such as people, organizations, and locations. The machine translation was used to convert the comments from one language to another.

한국어 문법에 의한 톨카피얌 해독의 가능성 검토

Tolkāpiyam, the oldest known book written in Old Tamil, is believed to have been authored around 100 B.C. Although Tolkāpiyam deals with the grammar of Tamil, a language still spoken by Tamils today, this ancient text, along with other Old Tamil books, has remained largely undeciphered. The reason for this lies in the fact that the grammar of Old Tamil has not been fully comprehended. Previous attempts to decipher Tolkāpiyam have mostly involved making educated guesses about the meaning of a paragraph based on the meanings of individual words.
 In this paper, I reveal the grammatical system of Old Tamil, drawing from the grammar theory presented in Part 2 of Tolkāpiyam. Old Tamil, much like pre-medieval Korean, employs predicates as nouns. It features elements such as chosa (조사), verb endings, honorifics, and causative verb suffixes. Remarkably, Tolkāpiyam is structured following the Korean word order. The grammar of Old Tamil closely resembles that of modern Korean. As a demonstration that Koreans can decipher Tolkāpiyam without prior knowledge of ancient Tamil, relying on word meanings alone, I also provide a correct interpretation for several paragraphs in Part I (Phonology). These interpretations challenge previous misconceptions and affirm the validity of my paper's argument, highlighting the limitations of prior research.

Improvement of post BER performance in computer communication on Viterbi algorithm

Computer communication is the pivotal exchange of information and data among computing devices, exerting substantial influence over network communication quality. The conventional approach has typically centred on enhancing Bit Error Rate (BER) performance to bolster information quality during the transmission process. However, our research pioneers a novel strategy, focusing on post-BER improvement. Our innovative method prioritizes enhancing information quality at the receiver's end by implementing a unique approach. We employ the Viterbi algorithm to elevate the quality of individual words, ensuring accurate word recognition and precise comprehension. The ultimate goal is to provide end-users with an unambiguous and accurate understanding of transmitted words. Our simulation results conclusively demonstrate the effectiveness of this approach. Following the reception of words, we successfully attain the correct word information with remarkable accuracy. This approach transforms word-level information processing, safeguarding against inaccuracies and ambiguities that may otherwise compromise the integrity of data transfer. In summary, our research introduces a paradigm shift in optimizing information quality within computer communication networks. By focusing on post-BER improvements and harnessing the power of the Viterbi algorithm, we enable the receiver to decode words with precision and clarity. This breakthrough promises to enhance the overall quality and reliability of information exchange, further solidifying the foundation of modern computer communication.

A novel self-supervised contrastive learning based sentence-level attribute induction method for online satisfaction evaluation

Satisfaction evaluation provides valuable support for procurement decisions and product improvement on e-commerce platforms. As an important task of satisfaction evaluation, attribute induction is essential to identify relevant aspects and demands of user-generated evaluations. However, previous methods have often directly predicted typical attributes using individual words, neglecting valuable sentence-level information that could significantly enhance effectiveness. To alleviate this issue, this article introduces the concept of representations of potential attributes (PARs) to explicitly capture sentence-level information regarding product attributes in customer reviews. It further develops a novel self-supervised contrastive learning based sentence-level attribute induction method (SSCL-SAI) to train these PARs. Recognizing that sentences are the basic units of human expression, this approach mitigates information loss. Then, an aspect-category sentiment analysis model is used to transform the unstructured online textual reviews into linguistic 2-tuple for obtaining final satisfaction evaluation result. These 2-tuples provide a representation of sentiment orientations on a continuous numerical scale, enabling precise linguistic operations. Afterwards, an illustrative example conducted on real-world laptop product satisfaction analysis testify the validity of the approach. Finally, the comparative analysis shows the effectiveness of the proposed SSCL-SAI, and experiments on the impact of hyperparameter settings demonstrate that SSCL-SAI can be applied in various scenarios.

Research advanced in Chinese word segmentation methods and challenges

Chinese word segmentation refers to the process of dividing a sequence of Chinese characters into individual words. It constitutes a fundamental component of Chinese natural language processing. Due to the intricacies of the Chinese language, Chinese word segmentation has garnered significant attention from researchers. Based on a review of historical literature, segmentation methods can be broadly categorized into rule-based, statistical, semantic-based, and comprehension-based approaches. With the advancement of machine learning, neural networks have emerged as the mainstream algorithm for word segmentation. However, Chinese presents several unique challenges, leading to segmentation results that are less effective compared to morphological analysis in languages like English. Moreover, word segmentation faces new challenges such as dependency on the quality and scale of corpora, as well as domain-specific segmentation in diverse fields. Addressing these emerging challenges will undoubtedly become a focal point in future research endeavors in this field. This review provides a comprehensive summary of existing methods, discusses the current state of Chinese word segmentation, and outlines directions for addressing the evolving complexities in the field. As Chinese language processing continues to advance, finding robust solutions for accurate word segmentation remains a critical area of research.

الاحتجاج بالقرآن الكريم في علوم اللغة العربية

In the Name of Allah, the Most Merciful, the Most Compassionate. With His help, we seek assistance, and in Him, we put our trust. He is the Most Merciful of all. May peace and blessings be upon our Prophet Muhammad, sent as a mercy to all worlds, and upon his family, companions, and spouse, with abundant salutations until the Day of Judgment. Every matter has a guiding principle, and similarly, every language has sources it relies upon and draws from. Undoubtedly, the most important sources of the Arabic language are the Quran, the Prophetic Hadiths, and the poetry and sayings of the Arabs. The use of the Quran as evidence holds significant importance, serving as a source from which all branches of Arabic studies have drawn, whether it be linguistic dictionaries, books of interpretation (Tafsir), grammar, rhetoric, literature, or other areas within the various sciences of Arabic. This is because the Quran is considered the primary source in all these fields. Linguists have used the Quran from two perspectives: As a source to validate the accuracy of a linguistic term, grammatical structure, or a specific dialect of the Arabic language. It is a crucial reference for determining the correctness of meanings related to the sciences of Arabic. Its benefit is vast, whether at the level of individual Quranic words or syntactic structures, even down to phonological aspects, where Quranic words act as guiding lights for the interpretation and clarification of many language terms. It helps in determining the meanings of numerous Arabic terms, as the Quran is rich in linguistic expressions. In the interpretation of the Quran itself, the primary approach to understanding the meaning of Quranic terms is through the interpretation of the Quran by the Quran.This study includes several key topics, such as the meaning and significance of using evidence, its sources, the importance of citing the Quran, Quranic readings, and how linguists have benefited from citing the Quran and its readings. It also discusses the difference between Quranic readings and the Quran as a whole, especially in terms of their use as evidence in the sciences of Arabic. With this, we ask for Allah's guidance. Our Lord, accept this from us Indeed, you are the Hearing, the Knowing. (Surah Al-Baqarah, 2:127).

Individual word representations dissociate from linguistic context along a cortical unimodal to heteromodal gradient.

Language comprehension involves multiple hierarchical processing stages across time, space, and levels of representation. When processing a word, the sensory input is transformed into increasingly abstract representations that need to be integrated with the linguistic context. Thus, language comprehension involves both input-driven as well as context-dependent processes. While neuroimaging research has traditionally focused on mapping individual brain regions to the distinct underlying processes, recent studies indicate that whole-brain distributed patterns of cortical activation might be highly relevant for cognitive functions, including language. One such pattern, based on resting-state connectivity, is the 'principal cortical gradient', which dissociates sensory from heteromodal brain regions. The present study investigated the extent to which this gradient provides an organizational principle underlying language function, using a multimodal neuroimaging dataset of functional magnetic resonance imaging (fMRI) and magnetoencephalography (MEG) recordings from 102 participants during sentence reading. We found that the brain response to individual representations of a word (word length, orthographic distance, and word frequency), which reflect visual; orthographic; and lexical properties, gradually increases towards the sensory end of the gradient. Although these properties showed opposite effect directions in fMRI and MEG, their association with the sensory end of the gradient was consistent across both neuroimaging modalities. In contrast, MEG revealed that properties reflecting a word's relation to its linguistic context (semantic similarity and position within the sentence) involve the heteromodal end of the gradient to a stronger extent. This dissociation between individual word and contextual properties was stable across earlier and later time windows during word presentation, indicating interactive processing of word representations and linguistic context at opposing ends of the principal gradient. To conclude, our findings indicate that the principal gradient underlies the organization of a range of linguistic representations while supporting a gradual distinction between context-independent and context-dependent representations. Furthermore, the gradient reveals convergent patterns across neuroimaging modalities (similar location along the gradient) in the presence of divergent responses (opposite effect directions).

Deanthroponymic Contaminants as Objects of the Language Game (Based on Material from Spanish-Language Media)

This article is devoted to the study of anthroponymic occasionalisms and contaminants – lexical innovations created on the basis of anthroponyms – used in the Spanish-language media as part of the communicative strategy and creative potential of the language game realized in the course of the dynamic process that is characteristic of modern Spanish socio-political discourse. The objectives of the article are: a) to identify models of this method of word formation, and b) to establish the types of relations that appear between the components of the abovementioned derived words. Anthroponyms selected by random sampling from internet publications were studied, considered as objects of the language game, on the basis of which, as a result of deliberate word making, occasionalisms– contaminated formations that express, in addition to nominative meaning, expressiveness and emotive colouring – are created. The following methods were used in our analysis of the material: component analysis of the word; the word-formation method of studying transformational techniques; contextual analysis, which allows us to identify the communicative parameters of the use of language units and the contextual features of the implementation of their meaning in the text; the descriptive-analytical method to identify the essential characteristics and features of the linguistic facts under consideration against the background of extralinguistic factors; and the classification method and the method of linguistic and stylistic analysis, focused on identifying the specifics of individual word usage due to the communicative and aesthetic tasks of the author of the text. The relevance of this study is due to the attention of linguists to axiological issues, the increasing regularity of the use of language play in the media space in particular, and contamination as one of the means of expression in the Spanish-language media at the present stage, using anthroponyms as an object. Anthroponymic contaminated formations are becoming more common and diverse, and there is a need to systematize their models and study their structure and semantics. As a result of the examination carried out on a certain array of initial data, semanticstylistic and axiological functions of contaminants – mainly defamatory, invective – are established in accordance with the pragmatic goals of the author of the text. The models of creating anthroponymic contaminants by both occasional and conventional means of word formation are identified and systematized.

Tiered vocabulary and raciolinguistic discourses of deficit: from academic scholarship to education policy

Tiered vocabulary is a pervasive concept in academic scholarship, education policy, and schools. It involves placing individual words into hierarchically arranged tiers, based on their apparent simplicity, sophistication, utility, and complexity, with these categorisations used to determine which words carry value in the classroom. In this article I conduct a genealogy of tiered vocabulary and argue that it is a raciolinguistic ideology which frames the language practices of racialised and working-class children as deficient and requiring modification. I trace the ideological roots of tiered vocabulary to European colonial representations of the purportedly limited vocabulary of Black African communities. I then examine how tiered vocabulary emerged as a concept in 1980s US academic knowledge production, based on experiments led by white academics on predominantly Black children from low-income homes. I show how tiered vocabulary is descendant from deficit, anti-Black thinking which characterised mainstream educational linguistics in the 1960s. I then focus on how it has become normalised in England’s education policy in the 2020s, as part of a flawed theory of change which deems that the acquisition of academic language is a tool for enabling social justice. Ultimately, I show how tiered vocabulary is a flawed theory of language which ties together race and class in producing discourses of linguistic deficiency, and legitimises language policing which undermines the education of marginalised children.

Individual word and phrase frequency effects in collocational processing: Evidence from typologically different languages, English and Turkish.

Collocations are understood to be integral building blocks of language processing, alongside individual words, but thus far evidence for the psychological reality of collocations has tended to be confined to English. In contrast to English, Turkish is an agglutinating language, utilizing productive morphology to convey complex meanings using a single word. Given this, we expected Turkish speakers to be less sensitive to phrasal frequencies than English speakers. In Study 1, we conducted a corpus analysis of translation-equivalent adjective-noun collocations (e.g., front door) and found differences between the two languages in frequency counts. In Study 2, we conducted a reaction time experiment to determine the sensitivity of native speakers of English and Turkish to the frequency of adjectives, nouns, and whole collocations. Turkish speakers were less sensitive to whole-phrase frequencies, as predicted, indicating that collocations are processed less holistically in Turkish than English. Both groups demonstrated that processing collocations involves combining information about individual words and phrases. Taken together, we show that speakers are sensitive to frequency information at multiple grain sizes that are attuned to the typology of different languages. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

Supplemental Material for Individual Word and Phrase Frequency Effects in Collocational Processing: Evidence From Typologically Different Languages, English and Turkish

Supplemental Material for Individual Word and Phrase Frequency Effects in Collocational Processing: Evidence From Typologically Different Languages, English and Turkish

Punctuation in English and Russian Languages

Punctuation or punctuation marks in English, as in Russian, is used to separate sentences, parts of complex sentences and words, to express intonation and semantic emphasis of individual words. In both languages, punctuation in the text is governed by certain rules, but in English these rules are less strict. Good punctuation is the ability to make the right choice in the right place so that the reader can best understand the author’s intonation. In the article the authors analyze the use of punctuation marks in English and Russian. Differences and similarities in the use of punctuation were also demonstrated in the article.

Explainable text-based features in predictive models of crowdfunding campaigns

AbstractReward-Based Crowdfunding offers an opportunity for innovative ventures that would not be supported through traditional financing. A key problem for those seeking funding is understanding which features of a crowdfunding campaign will sway the decisions of a sufficient number of funders. Predictive models of fund-raising campaigns used in combination with Explainable AI methods promise to provide such insights. However, previous work on Explainable AI has largely focused on quantitative structured data. In this study, our aim is to construct explainable models of human decisions based on analysis of natural language text, thus contributing to a fast-growing body of research on the use of Explainable AI for text analytics. We propose a novel method to construct predictions based on text via semantic clustering of sentences, which, compared with traditional methods using individual words and phrases, allows complex meaning contained in the text to be operationalised. Using experimental evaluation, we compare our proposed method to keyword extraction and topic modelling, which have traditionally been used in similar applications. Our results demonstrate that the sentence clustering method produces features with significant predictive power, compared to keyword-based methods and topic models, but which are much easier to interpret for human raters. We furthermore conduct a SHAP analysis of the models incorporating sentence clusters, demonstrating concrete insights into the types of natural language content that influence the outcome of crowdfunding campaigns.