We present Enterprise Alexandria, a new system for automatically constructing a knowledge base with high-precision and typed entities from private enterprise data such as emails, documents and intranet pages. Built as an extension of Alexandria [Winn et al., 2019], the key novelty of Enterprise Alexandria is the ability in processing both the textual information and the structured metadata available in each document in an online learning fashion, making use of any manual curations that have happened in the interim. This task is performed entirely eyes-off to respect the privacy of the user and the restricted access their documents. The knowledge discovery process uses a probabilistic program defining the process of generating the data item from a set of unknown typed entities. Using probabilistic inference, Enterprise Alexandria can jointly discover a large set of entities with custom types specific to the organization. Experiments on three real-world datasets show that the system outperforms alternative methods with the ability to work effectively at large scale.

Creating a knowledge base that is accurate, up-to-date and complete remains a significant challenge despite substantial efforts in automated knowledge base construction. In this paper, we present Alexandria -- a system for unsupervised, high-precision knowledge base construction. Alexandria uses a probabilistic program to define a process of converting knowledge base facts into unstructured text. Using probabilistic inference, we can invert this program and so retrieve facts, schemas and entities from web text. The use of a probabilistic program allows uncertainty in the text to be propagated through to the retrieved facts, which increases accuracy and helps merge facts from multiple sources. Because Alexandria does not require labelled training data, knowledge bases can be constructed with the minimum of manual input. We demonstrate this by constructing a high precision (typically 97\%+) knowledge base for people from a single seed fact.

Many aspects of the design of efficient crowdsourcing processes, such as defining worker's bonuses, fair prices and time limits of the tasks, involve knowledge of the likely duration of the task at hand. In this work we introduce a new time-sensitive Bayesian aggregation method that simultaneously estimates a task's duration and obtains reliable aggregations of crowdsourced judgments. Our method, called BCCTime, uses latent variables to represent the uncertainty about the workers' completion time, the tasks' duration and the workers' accuracy. To relate the quality of a judgment to the time a worker spends on a task, our model assumes that each task is completed within a latent time window within which all workers with a propensity to genuinely attempt the labelling task (i.e., no spammers) are expected to submit their judgments. In contrast, workers with a lower propensity to valid labelling, such as spammers, bots or lazy labellers, are assumed to perform tasks considerably faster or slower than the time required by normal workers. Specifically, we use efficient message-passing Bayesian inference to learn approximate posterior probabilities of (i) the confusion matrix of each worker, (ii) the propensity to valid labelling of each worker, (iii) the unbiased duration of each task and (iv) the true label of each task. Using two real-world public datasets for entity linking tasks, we show that BCCTime produces up to 11% more accurate classifications and up to 100% more informative estimates of a task's duration compared to state-of-the-art methods.

Crowdsourcing using active learning is described, for example, to select pairs of tasks and groups of workers so that information gained about answers to the tasks in the pool is optimized. In various examples a machine learning system learns variables describing characteristics of communities of workers, characteristics of workers, task variables and uncertainty of these variables. In various examples, the machine learning system predicts task variables and uncertainty of the predicted task variables for possible combinations of communities of workers and tasks. In examples the predicted variables and uncertainty are used to calculate expected information gain of the possible combinations and to rank the possible combinations. In examples, the crowdsourcing system uses the expected information gain to allocate tasks to worker communities and observe the results; the results may then be used to update the machine learning system.

We propose a new kind of probabilistic programming language for machine learning. We write programs simply by annotating existing relational schemas with probabilistic model expressions. We describe a detailed design of our language, Tabular, complete with formal semantics and type system. A rich series of examples illustrates the expressiveness of Tabular. We report an implementation, and show evidence of the succinctness of our notation relative to current best practice. Finally, we describe and verify a transformation of Tabular schemas so as to predict missing values in a concrete database. The ability to query for missing values provides a uniform interface to a wide variety of tasks, including classification, clustering, recommendation, and ranking.

This paper addresses the problem of extracting accurate labels from crowdsourced datasets, a key challenge in crowdsourcing. Prior work has focused on modeling the reliability of individual workers, for instance, by way of confusion matrices, and using these latent traits to estimate the true labels more accurately. However, this strategy becomes ineffective when there are too few labels per worker to reliably estimate their quality. To mitigate this issue, we propose a novel community-based Bayesian label aggregation model, CommunityBCC, which assumes that crowd workers conform to a few different types, where each type represents a group of workers with similar confusion matrices. We assume that each worker belongs to a certain community, where the worker's confusion matrix is similar to (a perturbation of) the community's confusion matrix. Our model can then learn a set of key latent features: (i) the confusion matrix of each community, (ii) the community membership of each user, and (iii) the aggregated label of each item. We compare the performance of our model against established aggregation methods on a number of large-scale, real-world crowdsourcing datasets. Our experimental results show that our CommunityBCC model consistently outperforms state-of-the-art label aggregation methods, requiring, on average, 50% less data to pass the 90% accuracy mark.