This paper reports our explorations on learning Sensory Media Association through Reciprocating Training (SMART). The proposed learning system contains two deep autoencoders, one for learning speech representations and another for learning image representations. Two deep networks are trained to bridge the latent spaces of two autoencoders, yielding representation mappings for both speech-to-image and image-to-speech. To improve feature clustering in both latent spaces, the system alternately uses one modality to guide the learning of another modality. Different from traditional technology that uses a fixed modality for supervision (e.g. using text labels for image classification), the proposed approach facilitates a machine to learn from sensory data of two or more modalities through alternating guidance among these modalities. We evaluate the proposed model with MNIST digit images and corresponding digit speeches in the Google Command Digit Dataset (GCDD). We also evaluate the model with a dataset based on COIL-100 and corresponding Watson synthesized speech. The results demonstrate the model's promising viability for sensory media association.

Automatic medical report generation from chest X-ray images is one possibility for assisting doctors to reduce their workload. However, the different patterns and data distribution of normal and abnormal cases can bias machine learning models. Previous attempts did not focus on isolating the generation of the abnormal and normal sentences in order to increase the variability of generated paragraphs. To address this, we propose to separate abnormal and normal sentence generation by using a dual word LSTM in a hierarchical LSTM model. In addition, we conduct an analysis on the distinctiveness of generated sentences compared to the BLEU score, which increases when less distinct reports are generated. Together with this analysis, we propose a way of selecting a model that generates more distinctive sentences. We hope our findings will help to encourage the development of new metrics to better verify methods of automatic medical report generation.

Twitter and other social media platforms are often used for sharing interest in products. The identification of purchase decision stages, such as in the AIDA model (Awareness, Interest, Desire, Action), can enable more personalized e-commerce services and a finer-grained targeting of ads than predicting purchase intent only. In this paper, we propose and analyze neural models for identifying the purchase stage of single tweets in a user's tweet sequence. In particular, we identify three challenges of purchase stage identification: imbalanced label distribution with a high number of negative instances, limited amount of training data, and domain adaptation with no or only little target domain data. Our experiments reveal that the imbalanced label distribution is the main challenge for our models. We address it with ranking loss and perform detailed investigations of the performance of our models on the different output classes. In order to improve the generalization of the models and augment the limited amount of training data, we examine the use of sentiment analysis as a complementary, secondary task in a multitask framework. For applying our models to tweets from another product domain, we consider two scenarios: For the first scenario without any labeled data in the target product domain, we show that learning domain-invariant representations with adversarial training is most promising while for the second scenario with a small number of labeled target examples, finetuning the source model weights performs best. Finally, we conduct several analyses, including extracting attention weights and representative phrases for the different purchase stages. The results suggest that the model is learning features indicative of purchase stages and that the confusion errors are sensible.

A common issue in training a deep learning, abstractive summarization model is lack of a large set of training summaries. This paper examines techniques for adapting from a labeled source domain to an unlabeled target domain in the context of an encoder-decoder model for text generation. In addition to adversarial domain adaptation (ADA), we introduce the use of artificial titles and sequential training to capture the grammatical style of the unlabeled target domain. Evaluation on adapting to/from news articles and Stack Exchange posts indicates that the use of these techniques can boost performance for both unsupervised adaptation as well as fine-tuning with limited target data.

In multi-participant postings, as in online chat conversations, several conversations or topic threads may take place concurrently. This leads to difficulties for readers reviewing the postings in not only following discussions but also in quickly identifying their essence. A two-step process, disentanglement of interleaved posts followed by summarization of each thread, addresses the issue, but disentanglement errors are propagated to the summarization step, degrading the overall performance. To address this, we propose an end-to-end trainable encoder-decoder network for summarizing interleaved posts. The interleaved posts are encoded hierarchically, i.e., word-to-word (words in a post) followed by post-to-post (posts in a channel). The decoder also generates summaries hierarchically, thread-to-thread (generate thread representations) followed by word-to-word (i.e., generate summary words). Additionally, we propose a hierarchical attention mechanism for interleaved text. Overall, our end-to-end trainable hierarchical framework enhances performance over a sequence to sequence framework by 8\% on a synthetic interleaved texts dataset.

We describe a corpus analysis method to extract terminology from a collection of technical specifications book in the field of construction. Using statistics and word n-grams analyzes, we extract the terminology of the domain and then perform pruning steps with linguistic patterns and internet queries to improve the quality of the final terminology. In this paper we specifically focus on the improvements got by applying Internet queries and patterns. These improvements are evaluated by using a manual evaluation carried out by 6 experts in the field in the case of technical specification books.

We leverage a popularity measure in social media as a distant label for extractive summarization of online conversations. In social media, users can vote, share, or bookmark a post they prefer. The number of these actions is regarded as a measure of popularity. However, popularity is not solely determined by content of a post, e.g., a text or an image in a post, but is highly contaminated by its contexts, e.g., timing, and authority. We propose a disjunctive model, which computes the contribution of content and context separately. For evaluation, we build a dataset where the informativeness of a comment is annotated. We evaluate the results with ranking metrics, and show that our model outperforms the baseline model, which directly uses popularity as a measure of informativeness.

An enormous amount of conversation occurs online every day, including on chat platforms where multiple conversations may take place concurrently. Interleaved conversations lead to difficulties in not only following discussions but also retrieving relevant information from simultaneous messages. Conversation disentanglement aims to separate overlapping messages into detached conversations. In this paper, we propose to leverage representation learning for conversation disentanglement. A Siamese Hierarchical Convolutional Neural Network (SHCNN), which integrates local and more global representations of a message, is first presented to estimate the conversation-level similarity between closely posted messages. With the estimated similarity scores, our algorithm for Conversation Identification by SImilarity Ranking (CISIR) then derives conversations based on high-confidence message pairs and pairwise redundancy. Experiments were conducted with four publicly available datasets of conversations from Reddit and IRC channels. The experimental results show that our approach significantly outperforms comparative baselines in both pairwise similarity estimation and conversation disentanglement.

Convolutional Neural Networks (CNN) have successfully been utilized for localization using a single monocular image [1]. Most of the work to date has either focused on reducing the dimensionality of data for better learning of parameters during training or on developing different variations of CNN models to improve pose estimation. Many of the best performing works solely consider the content in a single image, while the context from historical images is ignored. In this paper, we propose a combined CNN-LSTM which is capable of incorporating contextual information from historical images to better estimate the current pose. Experimental results achieved using a dataset collected in an indoor office space improved the overall system results to 0.8 m & 2.5° at the third quartile of the cumulative distribution as compared with 1.5 m & 3.0° achieved by PoseNet [1]. Furthermore, we demonstrate how the temporal information exploited by the CNN-LSTM model assists in localizing the robot in situations where image content does not have sufficient features.

Video summarization and video captioning are considered two separate tasks in existing studies. For longer videos, automatically identifying the important parts of video content and annotating them with captions will enable a richer and more concise condensation of the video. We propose a general neural network architecture that jointly considers two supervisory signals (i.e., an image-based video summary and text-based video captions) in the training phase and generates both a video summary and corresponding captions for a given video in the test phase. Our main idea is that the summary signals can help a video captioning model learn to focus on important frames. On the other hand, caption signals can help a video summarization model to learn better semantic representations. Jointly modeling both the video summarization and the video captioning tasks offers a novel end-to-end solution that generates a captioned video summary enabling users to index and navigate through the highlights in a video. Moreover, our experiments show the joint model can achieve better performance than state-of- the-art approaches in both individual tasks.

The availability of mobile access has shifted social media use. With that phenomenon, what users shared on social media and where they visited is naturally an excellent resource to learn their visiting behavior. Knowing visit behaviors would help market survey and customer relationship management, e.g., sending customers coupons of the businesses that they visit frequently. Most prior studies leverage meta-data e.g., check- in locations to profile visiting behavior but neglect important information from user-contributed content, e.g., images. This work addresses a novel use of image content for predicting the user visit behavior, i.e., the frequent and regular business venue categories that the content owner would visit. To collect training data, we propose a strategy to use geo-metadata associated with images for deriving the labels of an image owner’s visit behavior. Moreover, we model a user’s sequential images by using an end-to-end learning framework to reduce the optimization loss. That helps improve the prediction accuracy against the baseline as demonstrated in our experiments. The prediction is completely based on image content that is more available in social media than geo-metadata, and thus allows coverage in profiling a wider set of users.

Users often use social media to share their interest in products. We propose to identify purchase stages from Twitter data following the AIDA model (Awareness, Interest, Desire, Action). In particular, we define a task of classifying the purchase stage of each tweet in a user's tweet sequence. We introduce RCRNN, a Ranking Convolutional Recurrent Neural Network which computes tweet representations using convolution over word embeddings and models a tweet sequence with gated recurrent units. Also, we consider various methods to cope with the imbalanced label distribution in our data and show that a ranking layer outperforms class weights.

The abundance of data posted to Twitter enables companies to extract useful information, such as Twitter users who are dissatisfied with a product. We endeavor to determine which Twitter users are potential customers for companies and would be receptive to product recommendations through the language they use in tweets after mentioning a product of interest. With Twitter's API, we collected tweets from users who tweeted about mobile devices or cameras. An expert annotator determined whether each tweet was relevant to customer purchase behavior and whether a user, based on their tweets, eventually bought the product. For the relevance task, among four models, a feed-forward neural network yielded the best cross-validation accuracy of over 80% per product. For customer purchase prediction of a product, we observed improved performance with the use of sequential input of tweets to recurrent models, with an LSTM model being best; we also observed the use of relevance predictions in our model to be more effective with less powerful RNNs and on more difficult tasks.

Captions are a central component in image posts that communicate the background story behind photos. Captions can enhance the engagement with audiences and are therefore critical to campaigns or advertisement. Previous studies in image captioning either rely solely on image content or summarize multiple web documents related to image's location; both neglect users' activities. We propose business-aware latent topics as a new contextual cue for image captioning that represent user activities. The idea is to learn the typical activities of people who posted images from business venues with similar categories (e.g., fast food restaurants) to provide appropriate context for similar topics (e.g., burger) in new posts. User activities are modeled via a latent topic representation. In turn, the image captioning model can generate sentences that better reflect user activities at business venues. In our experiments, the business-aware latent topics are effective for adapting to captions to images captured in various businesses than the existing baselines. Moreover, they complement other contextual cues (image, time) in a multi-modal framework.

Many people post about their daily life on social media. These posts may include information about the purchase activity of people, and insights useful to companies can be derived from them: e.g. profile information of a user who mentioned something about their product. As a further advanced analysis, we consider extracting users who are likely to buy a product from the set of users who mentioned that the product is attractive. In this paper, we report our methodology for building a corpus for Twitter user purchase behavior prediction. First, we collected Twitter users who posted a want phrase + product name: e.g. "want a Xperia" as candidate want users, and also candidate bought users in the same way. Then, we asked an annotator to judge whether a candidate user actually bought a product. We also annotated whether tweets randomly sampled from want/bought user timelines are relevant or not to purchase. In this annotation, 58% of want user tweets and 35% of bought user tweets were annotated as relevant. Our data indicate that information embedded in timeline tweets can be used to predict purchase behavior of tweeted products.

Image localization is important for marketing and recommendation of local business; however, the level of granularity is still a critical issue. Given a consumer photo and its rough GPS information, we are interested in extracting the fine-grained location information (i.e. business venues) of the image. To this end, we propose a novel framework for business venue recognition. The framework mainly contains three parts. First, business aware visual concept discovery: we mine a set of concepts that are useful for business venue recognition based on three guidelines including business-awareness, visually detectable, and discriminative power. Second, business-aware concept detection by convolutional neural networks (BA-CNN): we pro- pose a new network architecture that can extract semantic concept features from input image. Third, multimodal business venue recognition: we extend visually detected concepts to multimodal feature representations that allow a test image to be associated with business reviews and images from social media for business venue recognition. The experiments results show the visual concepts detected by BA-CNN can achieve up to 22.5% relative improvement for business venue recognition compared to the state-of-the-art convolutional neural network features. Experiments also show that by leveraging multimodal information from social media we can further boost the performance, especially in the case when the database images belonging to each business venue are scarce.

In this paper, we analyze the association between a social media user's photo content and their interests. Visual content of photos is analyzed using state-of-the-art deep learning based automatic concept recognition. An aggregate visual concept signature is thereby computed for each user. User tags manually applied to their photos are also used to construct a tf-idf based signature per user. We also obtain social groups that users join to represent their social interests. In an effort to compare the visual-based versus tag-based user profiles with social interests, we compare corresponding similarity matrices with a reference similarity matrix based on users' group memberships. A random baseline is also included that groups users by random sampling while preserving the actual group sizes. A difference metric is proposed and it is shown that the combination of visual and text features better approximates the group-based similarity matrix than either modality individually. We also validate the visual analysis against the reference inter-user similarity using the Spearman rank correlation coefficient. Finally we cluster users by their visual signatures and rank clusters using a cluster uniqueness criteria.

We present a general framework and working system for predicting likely affective responses of the viewers in the social media environment after an image is posted online. Our approach emphasizes a mid-level concept representation, in which intended affects of the image publisher is characterized by a large pool of visual concepts (termed PACs) detected from image content directly instead of textual metadata, evoked viewer affects are represented by concepts (termed VACs) mined from online comments, and statistical methods are used to model the correlations among these two types of concepts. We demonstrate the utilities of such approaches by developing an end-to-end Assistive Comment Robot application, which further includes components for multi-sentence comment generation, interactive interfaces, and relevance feedback functions. Through user studies, we showed machine suggested comments were accepted by users for online posting in 90% of completed user sessions, while very favorable results were also observed in various dimensions (plausibility, preference, and realism) when assessing the quality of the generated image comments.

Name of an identity is strongly influenced by his/her cultural background such as gender and ethnicity, both vital attributes for user profiling, attribute-based retrieval, etc. Typically, the associations between names and attributes (e.g., people named "Amy" are mostly females) are annotated manually or provided by the census data of governments. We propose to associate a name and its likely demographic attributes by exploiting click-throughs between name queries and images with automatically detected facial attributes. This is the first work attempting to translate an abstract name to demographic attributes in visual-data-driven manner, and it is adaptive to incremental data, more countries and even unseen names (the names out of click-through data) without additional manual labels. In the experiments, the automatic name-attribute associations can help gender inference with competitive accuracy by using manual labeling. It also benefits profiling social media users and keyword-based face image retrieval, especially for contributing 12% relative improvement of accuracy in adapting to unseen names.