Chandra Bhagavatula, Sergey Feldman, Russell Power, Waleed Ammar. Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers). 2018.

Pre-trained word embeddings learned from unlabeled text have become a standard component of neural network architectures for NLP tasks.However, in most cases, the recurrent network that operates on word-level representations to produce context sensitive representations is trained on relatively little labeled data.In this paper, we demonstrate a general semi-supervised approach for adding pretrained context embeddings from bidirectional language models to NLP systems and apply it to sequence labeling tasks.We evaluate our model on two standard datasets for named entity recognition (NER) and chunking, and in both cases achieve state of the art results, surpassing previous systems that use other forms of transfer or joint learning with additional labeled data and task specific gazetteers.

This paper proposes K-NRM, a kernel based neural model for document ranking. Given a query and a set of documents, K-NRM uses a translation matrix that models word-level similarities via word embeddings, a new kernel-pooling technique that uses kernels to extract multi-level soft match features, and a learning-to-rank layer that combines those features into the final ranking score. The whole model is trained end-to-end. The ranking layer learns desired feature patterns from the pairwise ranking loss. The kernels transfer the feature patterns into soft-match targets at each similarity level and enforce them on the translation matrix. The word embeddings are tuned accordingly so that they can produce the desired soft matches. Experiments on a commercial search engine's query log demonstrate the improvements of K-NRM over prior feature-based and neural-based states-of-the-art, and explain the source of K-NRM's advantage: Its kernel-guided embedding encodes a similarity metric tailored for matching query words to document words, and provides effective multi-level soft matches.

This paper introduces Explicit Semantic Ranking (ESR), a new ranking technique that leverages knowledge graph embedding. Analysis of the query log from our academic search engine, SemanticScholar.org, reveals that a major error source is its inability to understand the meaning of research concepts in queries. To addresses this challenge, ESR represents queries and documents in the entity space and ranks them based on their semantic connections from their knowledge graph embedding. Experiments demonstrate ESR's ability in improving Semantic Scholar's online production system, especially on hard queries where word-based ranking fails.

Currently, users of geo-distributed storage systems face a hard choice between having serializable transactions with high latency, or limited or no transactions with low latency. We show that it is possible to obtain both serializable transactions and low latency, under two conditions. First, transactions are known ahead of time, permitting an a priori static analysis of conflicts. Second, transactions are structured as transaction chains consisting of a sequence of hops, each hop modifying data at one server. To demonstrate this idea, we built Lynx, a geo-distributed storage system that offers transaction chains, secondary indexes, materialized join views, and geo-replication. Lynx uses static analysis to determine if each hop can execute separately while preserving serializability---if so, a client needs wait only for the first hop to complete, which occurs quickly. To evaluate Lynx, we built three applications: an auction service, a Twitter-like microblogging site and a social networking site. These applications successfully use chains to achieve low latency operation and good throughput.

We describe the design and implementation of Walter, a key-value store that supports transactions and replicates data across distant sites. A key feature behind Walter is a new property called Parallel Snapshot Isolation (PSI). PSI allows Walter to replicate data asynchronously, while providing strong guarantees within each site. PSI precludes write-write conflicts, so that developers need not worry about conflict-resolution logic. To prevent write-write conflicts and implement PSI, Walter uses two new and simple techniques: preferred sites and counting sets. We use Walter to build a social networking application and port a Twitter-like application.