Named entity recognition (NER) is an important task in NLP, which is all the more challenging in conversational domain with their noisy facets. Moreover, conversational texts are often available in limited amount, making supervised tasks infeasible. To learn from small data, strong inductive biases are required. Previous work relied on hand-crafted features to encode these biases until transfer learning emerges. Here, we explore a transfer learning method, namely language model pretraining, on NER task in Indonesian conversational texts. We utilize large unlabeled data (generic domain) to be transferred to conversational texts, enabling supervised training on limited in-domain data. We report two transfer learning variants, namely supervised model fine-tuning and unsupervised pretrained LM fine-tuning. Our experiments show that both variants outperform baseline neural models when trained on small data (100 sentences), yielding an absolute improvement of 32 points of test F1 score. Furthermore, we find that the pretrained LM encodes part-of-speech information which is a strong predictor for NER.
We investigate the behaviour of attention in neural models of visually grounded speech trained on two languages: English and Japanese. Experimental results show that attention focuses on nouns and this behaviour holds true for two very typologically different languages. We also draw parallels between artificial neural attention and human attention and show that neural attention focuses on word endings as it has been theorised for human attention. Finally, we investigate how two visually grounded monolingual models can be used to perform cross-lingual speech-to-speech retrieval. For both languages, the enriched bilingual (speech-image) corpora with part-of-speech tags and forced alignments are distributed to the community for reproducible research.
The Generative Adversarial Network (GAN) has recently been applied to generate synthetic images from text. Despite significant advances, most current state-of-the-art algorithms are regular-grid region based; when attention is used, it is mainly applied between individual regular-grid regions and a word. These approaches are sufficient to generate images that contain a single object in its foreground, such as a "bird" or "flower". However, natural languages often involve complex foreground objects and the background may also constitute a variable portion of the generated image. Therefore, the regular-grid based image attention weights may not necessarily concentrate on the intended foreground region(s), which in turn, results in an unnatural looking image. Additionally, individual words such as "a", "blue" and "shirt" do not necessarily provide a full visual context unless they are applied together. For this reason, in our paper, we proposed a novel method in which we introduced an additional set of attentions between true-grid regions and word phrases. The true-grid region is derived using a set of auxiliary bounding boxes. These auxiliary bounding boxes serve as superior location indicators to where the alignment and attention should be drawn with the word phrases. Word phrases are derived from analysing Part-of-Speech (POS) results. We perform experiments on this novel network architecture using the Microsoft Common Objects in Context (MSCOCO) dataset and the model generates $256 \times 256$ conditioned on a short sentence description. Our proposed approach is capable of generating more realistic images compared with the current state-of-the-art algorithms.