Abstract not found

Topic models aid analysis of text corpora by identifying la-tent topics based on co-occurring words. Real-world de-ployments of topic models, however, often require intensive expert verification and model refinement. In this paper we present Termite, a visual analysis tool for assessing topic model quality. Termite uses a tabular layout to promote comparison of terms both within and across latent topics. We contribute a novel saliency measure for selecting relevant terms and a seriation algorithm that both reveals clustering structure and promotes the legibility of related terms. In a series of examples, we demonstrate how Termite allows analysts to identify coherent and significant themes.

Please do not redistribute or quote By studying the changes in employees ’ networks and performance before and after the introduction of a social networking tool, I find that a structurally diverse network (low in cohesion and rich in structural holes) has a positive effect on work performance. The size of the effect is smaller than traditional estimates, suggesting that omitted individual characteristics may bias the estimated network effect. I consider two intermediate mechanisms by which a structurally diverse network is theorized to improve work performance: information diversity (instrumental) and social communication (expressive) and quantify their effects on two types of work outcomes: billable revenue and layoffs. Analysis shows that the information diversity derived from a structurally diverse network is more correlated with generating billable revenue than is social communication. However, the opposite is true for layoffs. Friendship, as approximated by social communications, is more correlated with reduced layoff risks than is information diversity. Field interviews suggest that friends can serve as advocates in critical situations, ensuring that favorable information is distributed to decision makers. This, in turn, suggests that having a structurally diverse network can drive both work performance and job security, but that there is a tradeoff between either mobilizing friendship or gathering diverse information. Furthermore, it is important to examine the mechanisms by which social communications reduce the risks of being laid off. If social communications promote team effectiveness, delegating decisions rights to managers is optimal. However, if managers choose to optimize their own power at the expense of the firm, the positive impact of social communications on layoffs is evidence that delegating layoff decisions to managers can incur important costs.

Translating an information need into a keyword query can be a complex cognitive process which often results in under-specification. Retrieving documents based solely on key-words can lead the user to browse documents that do not address the specific query facets she was looking for. We introduce an unsupervised method for mining and modeling latent search concepts in order to increase the coverage of these facets. We use Latent Dirichlet Allocation (LDA), a generative probabilistic topic model, to exhibit highly-specific query-related topics from pseudo-relevant feedback documents. We define these topics as the latent concepts of the user query. The main strength of our approach is that it automatically estimates the number of latent con-cepts as well as the needed amount of feedback documents, without any prior training step. We evaluate our approach over two large ad-hoc TREC collections, and results show that our approach significantly improves document retrieval effectiveness and even provides a better representation of the information need than the original query.

Copyright by

Abstract not found