Karl Stratos

Understanding and predicting importance in images

What do people care about in an image? To drive computational visual recognition toward more human-centric outputs, we need a better understanding of how people perceive and judge the importance of content in images. In this paper, we explore how a number of factors relate to human perception of importance. Proposed factors fall into 3 broad types: 1) factors related to composition, e.g. size, location, 2) factors related to semantics, e.g. category of object or scene, and 3) contextual factors related to the likelihood of attribute-object, or object-scene pairs. We explore these factors using what people describe as a proxy for importance. Finally, we build models to predict what will be described about an image given either known image content, or image content estimated automatically by recognition systems.

Model-based Word Embeddings from Decompositions of Count Matrices

This work develops a new statistical understanding of word embeddings induced from transformed count data. Using the class of hidden Markov models (HMMs) underlying Brown clustering as a generative model, we demonstrate how canonical correlation analysis (CCA) and certain count transformations permit efficient and effective recovery of model parameters with lexical semantics. We further show in experiments that these techniques empirically outperform existing spectral methods on word similarity and analogy tasks, and are also competitive with other popular methods such as WORD2VEC and GLOVE.

New Transfer Learning Techniques for Disparate Label Sets

In natural language understanding (NLU), a user utterance can be labeled differently depending on the domain or application (e.g., weather vs. calendar). Standard domain adaptation techniques are not directly applicable to take advantage of the existing annotations because they assume that the label set is invariant. We propose a solution based on label embeddings induced from canonical correlation analysis (CCA) that reduces the problem to a standard domain adaptation task and allows use of a number of transfer learning techniques. We also introduce a new transfer learning technique based on pretraining of hidden-unit CRFs (HUCRFs). We perform extensive experiments on slot tagging on eight personal digital assistant domains and demonstrate that the proposed methods are superior to strong baselines.

Weakly Supervised Slot Tagging with Partially Labeled Sequences from Web Search Click Logs

In this paper, we apply a weakly-supervised learning approach for slot tagging using conditional random fields by exploiting web search click logs. We extend the constrained lattice training of T¨¨ om et al. (2013) to non-linear conditional random fields in which latent variables mediate between observations and labels. When combined with a novel initialization scheme that leverages unlabeled data, we show that our method gives significant improvement over strong supervised and weakly-supervised baselines.

Large Scale Retrieval and Generation of Image Descriptions

What is the story of an image? What is the relationship between pictures, language, and information we can extract using state of the art computational recognition systems? In an attempt to address both of these questions, we explore methods for retrieving and generating natural language descriptions for images. Ideally, we would like our generated textual descriptions (captions) to both sound like a person wrote them, and also remain true to the image content. To do this we develop data-driven approaches for image description generation, using retrieval-based techniques to gather either: (a) whole captions associated with a visually similar image, or (b) relevant bits of text (phrases) from a large collection of image + description pairs. In the case of (b), we develop optimization algorithms to merge the retrieved phrases into valid natural language sentences. The end result is two simple, but effective, methods for harnessing the power of big data to produce image captions that are altogether more general, relevant, and human-like than previous attempts.

Adversarial Adaptation of Synthetic or Stale Data.

Two types of data shift common in practice are 1. transferring from synthetic data to live user data (a deployment shift), and 2. transferring from stale data to current data (a temporal shift). Both cause a distribution mismatch between training and evaluation, leading to a model that overfits the flawed training data and performs poorly on the test data. We propose a solution to this mismatch problem by framing it as domain adaptation, treating the flawed training dataset as a source domain and the evaluation dataset as a target domain. To this end, we use and build on several recent advances in neural domain adaptation such as adversarial training (Ganinet al., 2016) and domain separation network (Bousmalis et al., 2016), proposing a new effective adversarial training scheme. In both supervised and unsupervised adaptation scenarios, our approach yields clear improvement over strong baselines.

Domain Attention with an Ensemble of Experts

An important problem in domain adaptation is to quickly generalize to a new domain with limited supervision given K existing domains. One approach is to retrain a global model across all K + 1 domains using standard techniques, for instance Daume III (2009). However, it is desirable to adapt without having to re-estimate a global model from scratch each time a new domain with potentially new intents and slots is added. We describe a solution based on attending an ensemble of domain experts. We assume K domain specific intent and slot models trained on respective domains. When given domain K + 1, our model uses a weighted combination of the K domain experts’ feedback along with its own opinion to make predictions on the new domain. In experiments, the model significantly outperforms baselines that do not use domain adaptation and also performs better than the full retraining approach.

Compact Lexicon Selection with Spectral Methods

In this paper, we introduce the task of selecting compact lexicon from large, noisy gazetteers. This scenario arises often in practice, in particular spoken language understanding (SLU). We propose a simple and effective solution based on matrix decomposition techniques: canonical correlation analysis (CCA) and rank-revealing QR (RRQR) factorization. CCA is first used to derive low-dimensional gazetteer embeddings from domain-specific search logs. Then RRQR is used to find a subset of these embeddings whose span approximates the entire lexicon space. Experiments on slot tagging show that our method yields a small set of lexicon entities with average relative error reduction of > 50% over randomly selected lexicon.

Scalable Semi-Supervised Query Classification Using Matrix Sketching

The enormous scale of unlabeled text available today necessitates scalable schemes for representation learning in natural language processing. For instance, in this paper we are interested in classifying the intent of a user query. While our labeled data is quite limited, we have access to virtually an unlimited amount of unlabeled queries, which could be used to induce useful representations: for instance by principal component analysis (PCA). However, it is prohibitive to even store the data in memory due to its sheer size, let alone apply conventional batch algorithms. In this work, we apply the recently proposed matrix sketching algorithm to entirely obviate the problem with scalability (Liberty, 2013). This algorithm approximates the data within a specified memory bound while preserving the covariance structure necessary for PCA. Using matrix sketching, we significantly improve the user intent classification accuracy by leveraging large amounts of unlabeled queries.

Pre-training of Hidden-Unit CRFs

In this paper, we apply the concept of pretraining to hidden-unit conditional random fields (HUCRFs) to enable learning on unlabeled data. We present a simple yet effective pre-training technique that learns to associate words with their clusters, which are obtained in an unsupervised manner. The learned parameters are then used to initialize the supervised learning process. We also propose a word clustering technique based on canonical correlation analysis (CCA) that is sensitive to multiple word senses, to further improve the accuracy within the proposed framework. We report consistent gains over standard conditional random fields (CRFs) and HUCRFs without pre-training in semantic tagging, named entity recognition (NER), and part-of-speech (POS) tagging tasks, which could indicate the task independent nature of the proposed technique.