CLIPTrans: Transferring Visual Knowledge with Pre-trained Models for Multimodal Machine Translation

There has been a growing interest in developing multimodal machine translation (MMT) systems that enhance neural machine translation (NMT) with visual knowledge. This problem setup involves using images as auxiliary information during training, and more recently, eliminating their use during inference. Towards this end, previous works face a challenge in training powerful MMT models from scratch due to the scarcity of annotated multilingual vision-language data, especially for low-resource languages. Simultaneously, there has been an influx of multilingual pre-trained models for NMT and multimodal pre-trained models for vision-language tasks, primarily in English, which have shown exceptional generalisation ability. However, these are not directly applicable to MMT since they do not provide aligned multimodal multilingual features for generative tasks. To alleviate this issue, instead of designing complex modules for MMT, we propose CLIPTrans, which simply adapts the independently pre-trained multimodal M-CLIP and the multilingual mBART. In order to align their embedding spaces, mBART is conditioned on the M-CLIP features by a prefix sequence generated through a lightweight mapping network. We train this in a two-stage pipeline which warms up the model with image captioning before the actual translation task. Through experiments, we demonstrate the merits of this framework and consequently push forward the state-of-the-art across standard benchmarks by an average of +2.67 BLEU. The code can be found at www.github.com/devaansh100/CLIPTrans.

CascadeXML: Rethinking Transformers for End-to-end Multi-resolution Training in Extreme Multi-label Classification

Extreme Multi-label Text Classification (XMC) involves learning a classifier that can assign an input with a subset of most relevant labels from millions of label choices. Recent approaches, such as XR-Transformer and LightXML, leverage a transformer instance to achieve state-of-the-art performance. However, in this process, these approaches need to make various trade-offs between performance and computational requirements. A major shortcoming, as compared to the Bi-LSTM based AttentionXML, is that they fail to keep separate feature representations for each resolution in a label tree. We thus propose CascadeXML, an end-to-end multi-resolution learning pipeline, which can harness the multi-layered architecture of a transformer model for attending to different label resolutions with separate feature representations. CascadeXML significantly outperforms all existing approaches with non-trivial gains obtained on benchmark datasets consisting of up to three million labels. Code for CascadeXML will be made publicly available at \url{https://github.com/xmc-aalto/cascadexml}.

Embedding Convolutions for Short Text Extreme Classification with Millions of Labels

Automatic annotation of short-text data to a large number of target labels, referred to as Short Text Extreme Classification, has recently found numerous applications in prediction of related searches and product recommendation tasks. The conventional usage of Convolutional Neural Network (CNN) to capture n-grams in text-classification relies heavily on uniformity in word-ordering and the presence of long input sequences to convolve over. However, this is missing in short and unstructured text sequences encountered in search and recommendation. In order to tackle this, we propose an orthogonal approach by recasting the convolution operation to capture coupled semantics along the embedding dimensions, and develop a word-order agnostic embedding enhancement module to deal with the lack of structure in such queries. Benefitting from the computational efficiency of the convolution operation, Embedding Convolutions, when applied on the enriched word embeddings, result in a light-weight and yet powerful encoder (InceptionXML) that is robust to the inherent lack of structure in short-text extreme classification. Towards scaling our model to problems with millions of labels, we also propose InceptionXML+, which addresses the shortcomings of the dynamic hard-negative mining framework in the recently proposed LightXML by improving the alignment between the label-shortlister and extreme classifier. On popular benchmark datasets, we empirically demonstrate that the proposed method outperforms state-of-the-art deep extreme classifiers such as Astec by an average of 5% and 8% on the P@k and propensity-scored PSP@k metrics respectively.