Search for: All records

Through a computational reading of the online book reviewing community LibraryThing, we examine the dynamics of a collaborative tagging system and learn how its users refine and redefine literary genres. LibraryThing tags are overlapping and multi-dimensional, created in a shared space by thousands of users, including readers, bookstore owners, and librarians. A common understanding of genre is that it relates to the content of books, but this resource allows us to view genre as an intersection of user communities and reader values and interests. We explore different methods of computational genre measurement within the open space of user-created tags. We measure overlap between books, tags, and users, and we also measure the homogeneity of communities associated with genre tags and correlate this homogeneity with reviewing behavior.Finally, by analyzing the text of reviews, we identify the thematic signatures of genres on LibraryThing, revealing similarities and differences between them. These measurements are intended to elucidate the genre conceptions of the users, not, as in prior work, to normalize the tags or enforce a hierarchy. We find that LibraryThing users make sense of genre through a variety of values and expectations, many of which fall outside common definitions and understandings of genre. 

Much of the progress in contemporary NLP has come from learning representations, such as masked language model (MLM) contextual embeddings, that turn challenging problems into simple classification tasks. But how do we quantify and explain this effect? We adapt general tools from computational learning theory to fit the specific characteristics of text datasets and present a method to evaluate the compatibility between representations and tasks. Even though many tasks can be easily solved with simple bag-of-words (BOW) representations, BOW does poorly on hard natural language inference tasks. For one such task we find that BOW cannot distinguish between real and randomized labelings, while pre-trained MLM representations show 72x greater distinction between real and random labelings than BOW. This method provides a calibrated, quantitative measure of the difficulty of a classification-based NLP task, enabling comparisons between representations without requiring empirical evaluations that may be sensitive to initializations and hyperparameters. The method provides a fresh perspective on the patterns in a dataset and the alignment of those patterns with specific labels. 

A common factor in bias measurement methods is the use of hand-curated seed lexicons, but there remains little guidance for their selection. We gather seeds used in prior work, documenting their common sources and rationales, and in case studies of three English-language corpora, we enumerate the different types of social biases and linguistic features that, once encoded in the seeds, can affect subsequent bias measurements. Seeds developed in one context are often re-used in other contexts, but documentation and evaluation remain necessary precursors to relying on seeds for sensitive measurements. 

One commonly recognized feature of the Ancient Greek corpus is that later texts frequently imitate and allude to model texts from earlier time periods, but analysis of this phenomenon is mostly done for specific author pairs based on close reading and highly visible instances of imitation. In this work, we use computational techniques to examine the similarity of a wide range of Ancient Greek authors, with a focus on similarity between authors writing many centuries apart. We represent texts and authors based on their usage of high-frequency words to capture author signatures rather than document topics and measure similarity using Jensen- Shannon Divergence. We then analyze author similarity across centuries, finding high similarity between specific authors and across the corpus that is not common to all languages. 

Images can give us insights into the contextual meanings of words, but current image-text grounding approaches require detailed annotations. Such granular annotation is rare, expensive, and unavailable in most domain-specific contexts. In contrast, unlabeled multi-image, multi-sentence documents are abundant. Can lexical grounding be learned from such documents, even though they have significant lexical and visual overlap? Working with a case study dataset of real estate listings, we demonstrate the challenge of distinguishing highly correlated grounded terms, such as “kitchen” and “bedroom”, and introduce metrics to assess this document similarity. We present a simple unsupervised clustering-based method that increases precision and recall beyond object detection and image tagging baselines when evaluated on labeled subsets of the dataset. The proposed method is particularly effective for local contextual meanings of a word, for example associating “granite” with countertops in the real estate dataset and with rocky landscapes in a Wikipedia dataset. 

Traditional methods for adding locally private noise to bag-of-words features overwhelm the true signal in the text data, removing the properties of sparsity and non-negativity often relied upon by distributional semantic models. We argue the formulation of limited-precision local privacy, which guarantees privacy between documents of less than a user-specified maximum distance, is a more appropriate framework for bag-of-words features. To reduce the number of features to which we must add random noise, we also compress word features before adding noise, then decompress those features before model inference. We test randomized methods of aggregation as well as methods informed by distributional properties of words. Applying LDA and LSA to synthetic and real data, we show that these approaches produce distributional models closer to those in the original data. 

Most previous work in unsupervised semantic modeling in the presence of metadata has assumed that our goal is to make latent dimensions more correlated with metadata, but in practice the exact opposite is often true. Some users want topic models that highlight differences between, for example, authors, but others seek more subtle connections across authors. We introduce three metrics for identifying topics that are highly correlated with metadata, and demonstrate that this problem affects between 30 and 50% of the topics in models trained on two real-world collections, regardless of the size of the model. We find that we can predict which words cause this phenomenon and that by selectively subsampling these words we dramatically reduce topic-metadata correlation, improve topic stability, and maintain or even improve model quality 

Duplicate documents are a pervasive problem in text datasets and can have a strong effect on unsupervised models. Methods to remove duplicate texts are typically heuristic or very expensive, so it is vital to know when and why they are needed. We measure the sensitivity of two latent semantic methods to the presence of different levels of document repetition. By artificially creating different forms of duplicate text we confirm several hypotheses about how repeated text impacts models. While a small amount of duplication is tolerable, substantial over-representation of subsets of the text may overwhelm meaningful topical patterns. 

Multimodal machine learning algorithms aim to learn visual-textual correspondences. Previous work suggests that concepts with concrete visual manifestations may be easier to learn than concepts with abstract ones. We give an algorithm for automatically computing the visual concreteness of words and topics within multimodal datasets. We apply the approach in four settings, ranging from image captions to images/text scraped from historical books. In addition to enabling explorations of concepts in multi-modal datasets, our concreteness scores predict the capacity of machine learning algorithms to learn textual/visual relationships. We find that 1) concrete concepts are indeed easier to learn; 2) the large number of algorithms we consider have similar failure cases; 3) the precise positive relationship between concreteness and performance varies between datasets. We conclude with recommendations for using concreteness scores to facilitate future multi- modal research.