RCV1: A New Benchmark Collection for Text Categorization Research

By David D. Lewis, Yiming Yang, Tony G. Rose, Fan Li. 2004

3,202 citations recorded by the source corpus.

Subfields: Natural Language Processing, Evaluation & Benchmarks

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Natural Language Processing

In Evaluation & Benchmarks

Citations in the corpus

Revisiting LSTM Networks for Semi-Supervised Text Classification via Mixed Objective Function (2019, methodology): “…(JZ15a) datasets are widely used for binary sentiment classiﬁcation of movie reviews and Amazon product reviews respectively while AG-News , DBpedia (Zhang, Zhao, and LeCun 2015), and RCV1 (Lewis et al. 2004) are for topic classiﬁcation of news articles, Wikipedia, and Reuters corpus respectively.”
Character-level Convolutional Networks for Text Classification (2015, background): “However, most open datasets for text classification are quite small, and large-scale datasets are splitted with a significantly smaller training set than testing [21].”
FEVER: a Large-scale Dataset for Fact Extraction and VERification (2018, background): “Progress on the aforementioned tasks, as well as others such as text classification, has benefited from the availability large-scale datasets (Lewis et al., 2004; Bowman et al., 2015; Rajpurkar et al., 2016).”
MLQA: Evaluating Cross-lingual Extractive Question Answering (2019, methodology): “Given recent progress in cross-lingual tasks such as document classiﬁcation (Lewis et al., 2004; Klementiev et al., 2012; Schwenk and Li, 2018), semantic role labelling (Akbik et al., 2015) and NLI (Conneau et al., 2018), we argue that while multilingual QA training data might be useful but not…”
Natural Language Processing (Almost) from Scratch (2011, methodology): “Firstly, they may have experienced difficulties because they train 50-dimensional embeddings for 269K distinct words using a comparatively small training set (RCV1, 37M words), unlikely to contain enough instances of the rare words.”
Hogwild: A Lock-Free Approach to Parallelizing Stochastic Gradient Descent (2011, methodology): “We tested our sparse SVM implementation on the Reuters RCV1 data set on the binary text classification task CCAT [19].”
A Baseline for Detecting Misclassified and Out-of-Distribution Examples in Neural Networks (2016, background): “The Reuters 8 (Lewis et al., 2004) dataset has eight different news subjects with nearly 8000 stories in total.”
Optimization Methods for Large-Scale Machine Learning (2016, neutral): “M L ] 1 5 Ju n 20 16”
SimLex-999: Evaluating Semantic Models With (Genuine) Similarity Estimation (2014, methodology): “Col-lobert and Weston (2008) train their models on 852 million words of text from a 2007 dump of Wikipedia and the RCV1 Corpus (Lewis et al., 2004) and use their embeddings to achieve state-of-the-art results on a variety of NLP tasks.”
Convolutional Neural Network Architectures for Matching Natural Language Sentences (2014, methodology): “mpletion This is an artiﬁcial task designed to elucidate how different matching models can capture the correspondence between two clauses within a sentence. Basically, we take a sentence from Reuters [12]with two “balanced” clauses (with 8˘28 words) divided by one comma, and use the ﬁrst clause as S”
On the challenges of learning with inference networks on sparse, high-dimensional data (2017, neutral): “Even after extensive training the inference network can fail to tightly approximate L ( x ; θ, ψ ∗ ), suggesting that there may be limitations to the power of generic amortized inference; (3) optimizing ψ ( x ) during training ameliorates under-ﬁtting and yields signiﬁcantly better generative models”
ExpertMatcher: Automating ML Model Selection for Clients using Hidden Representations (2019, methodology): “…our proposed method on six datasets, such as objects, text, digits, biological and sensor, namely STl-10 (Coates et al. , 2011), MNIST (LeCun et al. , 1998), HAR (Anguita et al. , 2013), Reuters (Lewis et al. , 2004), Non Line of Sight (Tancik et al. , 2018) and Diabetic Retinopathy (Graham, 2015).”
Decentralized Stochastic Optimization and Gossip Algorithms with Compressed Communication (2019, methodology): “In the experiments we rely on the epsilon (Sonnenburg et al., 2008) and rcv 1 (Lewis et al., 2004) datasets (cf. Table 2).”
Distributed delayed stochastic optimization (2011, methodology): “(8) We use the Reuters RCV1 dataset [11], which consists of N ≈ 800000 news articles, each labeled with a combination of the four labels economics, government, commerce, and medicine.”
Sparse Partially Linear Additive Models (2014, methodology): “We use Reuters Corpus Volume I (RCV1) which involves binary classification (Lewis et al., 2004).”