Jason Weston

WSABIE: scaling up to large vocabulary image annotation

Image annotation datasets are becoming larger and larger, with tens of millions of images and tens of thousands of possible annotations. We propose a strongly performing method that scales to such datasets by simultaneously learning to optimize precision at the top of the ranked list of annotations for a given image and learning a low-dimensional joint embedding space for both images and annotations. Our method, called WSABIE, both outperforms several baseline methods and is faster and consumes less memory.

Large scale image annotation: learning to rank with joint word-image embeddings

Image annotation datasets are becoming larger and larger, with tens of millions of images and tens of thousands of possible annotations. We propose a strongly performing method that scales to such datasets by simultaneously learning to optimize precision at k of the ranked list of annotations for a given image and learning a low-dimensional joint embedding space for both images and annotations. Our method both outperforms several baseline methods and, in comparison to them, is faster and consumes less memory. We also demonstrate how our method learns an interpretable model, where annotations with alternate spellings or even languages are close in the embedding space. Hence, even when our model does not predict the exact annotation given by a human labeler, it often predicts similar annotations, a fact that we try to quantify by measuring the newly introduced “sibling” precision metric, where our method also obtains excellent results.

A semantic matching energy function for learning with multi-relational data

Large-scale relational learning becomes crucial for handling the huge amounts of structured data generated daily in many application domains ranging from computational biology or information retrieval, to natural language processing. In this paper, we present a new neural network architecture designed to embed multi-relational graphs into a flexible continuous vector space in which the original data is kept and enhanced. The network is trained to encode the semantics of these graphs in order to assign high probabilities to plausible components. We empirically show that it reaches competitive performance in link prediction on standard datasets from the literature as well as on data from a real-world knowledge base (WordNet). In addition, we present how our method can be applied to perform word-sense disambiguation in a context of open-text semantic parsing, where the goal is to learn to assign a structured meaning representation to almost any sentence of free text, demonstrating that it can scale up to tens of thousands of nodes and thousands of types of relation.

Open Question Answering with Weakly Supervised Embedding Models

Building computers able to answer questions on any subject is a long standing goal of artificial intelligence. Promising progress has recently been achieved by methods that learn to map questions to logical forms or database queries. Such approaches can be effective but at the cost of either large amounts of human-labeled data or by defining lexicons and grammars tailored by practitioners. In this paper, we instead take the radical approach of learning to map questions to vectorial feature representations. By mapping answers into the same space one can query any knowledge base independent of its schema, without requiring any grammar or lexicon. Our method is trained with a new optimization procedure combining stochastic gradient descent followed by a fine-tuning step using the weak supervision provided by blending automatically and collaboratively generated resources. We empirically demonstrate that our model can capture meaningful signals from its noisy supervision leading to major improvements over paralex, the only existing method able to be trained on similar weakly labeled data.

Semi-supervised multi-task learning for predicting interactions between HIV-1 and human proteins

Motivation: Protein–protein interactions (PPIs) are critical for virtually every biological function. Recently, researchers suggested to use supervised learning for the task of classifying pairs of proteins as interacting or not. However, its performance is largely restricted by the availability of truly interacting proteins (labeled). Meanwhile, there exists a considerable amount of protein pairs where an association appears between two partners, but not enough experimental evidence to support it as a direct interaction (partially labeled). Results: We propose a semi-supervised multi-task framework for predicting PPIs from not only labeled, but also partially labeled reference sets. The basic idea is to perform multi-task learning on a supervised classification task and a semi-supervised auxiliary task. The supervised classifier trains a multi-layer perceptron network for PPI predictions from labeled examples. The semi-supervised auxiliary task shares network layers of the supervised classifier and trains with partially labeled examples. Semi-supervision could be utilized in multiple ways. We tried three approaches in this article, (i) classification (to distinguish partial positives with negatives); (ii) ranking (to rate partial positive more likely than negatives); (iii) embedding (to make data clusters get similar labels). We applied this framework to improve the identification of interacting pairs between HIV-1 and human proteins. Our method improved upon the state-of-the-art method for this task indicating the benefits of semi-supervised multi-task learning using auxiliary information. Availability: http://www.cs.cmu.edu/∼qyj/HIVsemi Contact: qyj@cs.cmu.edu

Semantic Frame Identification with Distributed Word Representations

We present a novel technique for semantic frame identification using distributed representations of predicates and their syntactic context; this technique leverages automatic syntactic parses and a generic set of word embeddings. Given labeled data annotated with frame-semantic parses, we learn a model that projects the set of word representations for the syntactic context around a predicate to a low dimensional representation. The latter is used for semantic frame identification; with a standard argument identification method inspired by prior work, we achieve state-ofthe-art results on FrameNet-style framesemantic analysis. Additionally, we report strong results on PropBank-style semantic role labeling in comparison to prior work.

Learning to rank recommendations with the k-order statistic loss

Making recommendations by learning to rank is becoming an increasingly studied area. Approaches that use stochastic gradient descent scale well to large collaborative filtering datasets, and it has been shown how to approximately optimize the mean rank, or more recently the top of the ranked list. In this work we present a family of loss functions, the k-order statistic loss, that includes these previous approaches as special cases, and also derives new ones that we show to be useful. In particular, we present (i) a new variant that more accurately optimizes precision at k, and (ii) a novel procedure of optimizing the mean maximum rank, which we hypothesize is useful to more accurately cover all of the users tastes. The general approach works by sampling N positive items, ordering them by the score assigned by the model, and then weighting the example as a function of this ordered set. Our approach is studied in two real-world systems, Google Music and YouTube video recommendations, where we obtain improvements for computable metrics, and in the YouTube case, increased user click through and watch duration when deployed live on www.youtube.com.

Multi-Tasking with Joint Semantic Spaces for Large-Scale Music Annotation and Retrieval

Abstract Music prediction tasks range from predicting tags given a song or clip of audio, predicting the name of the artist, or predicting related songs given a song, clip, artist name or tag. That is, we are interested in every semantic relationship between the different musical concepts in our database. In realistically sized databases, the number of songs is measured in the hundreds of thousands or more, and the number of artists in the tens of thousands or more, providing a considerable challenge to standard machine learning techniques. In this work, we propose a method that scales to such datasets which attempts to capture the semantic similarities between the database items by modelling audio, artist names, and tags in a single low-dimensional semantic embedding space. This choice of space is learnt by optimizing the set of prediction tasks of interest jointly using multi-task learning. Our single model learnt by training on the joint objective function is shown experimentally to have improved accuracy over training on each task alone. Our method also outperforms the baseline methods tried and, in comparison to them, is faster and consumes less memory. We also demonstrate how our method learns an interpretable model, where the semantic space captures well the similarities of interest.

Nonlinear latent factorization by embedding multiple user interests

Classical matrix factorization approaches to collaborative filtering learn a latent vector for each user and each item, and recommendations are scored via the similarity between two such vectors, which are of the same dimension. In this work, we are motivated by the intuition that a user is a much more complicated entity than any single item, and cannot be well described by the same representation. Hence, the variety of a users interests could be better captured by a more complex representation. We propose to model the user with a richer set of functions, specifically via a set of latent vectors, where each vector captures one of the users latent interests or tastes. The overall recommendation model is then nonlinear where the matching score between a user and a given item is the maximum matching score over each of the users latent interests with respect to the items latent representation. We describe a simple, general and efficient algorithm for learning such a model, and apply it to large scale, real-world datasets from YouTube and Google Music, where our approach outperforms existing techniques.

A unified multitask architecture for predicting local protein properties.

A variety of functionally important protein properties, such as secondary structure, transmembrane topology and solvent accessibility, can be encoded as a labeling of amino acids. Indeed, the prediction of such properties from the primary amino acid sequence is one of the core projects of computational biology. Accordingly, a panoply of approaches have been developed for predicting such properties; however, most such approaches focus on solving a single task at a time. Motivated by recent, successful work in natural language processing, we propose to use multitask learning to train a single, joint model that exploits the dependencies among these various labeling tasks. We describe a deep neural network architecture that, given a protein sequence, outputs a host of predicted local properties, including secondary structure, solvent accessibility, transmembrane topology, signal peptides and DNA-binding residues. The network is trained jointly on all these tasks in a supervised fashion, augmented with a novel form of semi-supervised learning in which the model is trained to distinguish between local patterns from natural and synthetic protein sequences. The task-independent architecture of the network obviates the need for task-specific feature engineering. We demonstrate that, for all of the tasks that we considered, our approach leads to statistically significant improvements in performance, relative to a single task neural network approach, and that the resulting model achieves state-of-the-art performance.