Automatically associating social media posts with topics is an important prerequisite for effective search and recommendation on many social media platforms. However, topic classification of such posts is quite challenging because of (a) a large topic space (b) short text with weak topical cues, and (c) multiple topic associations per post. In contrast to most prior work which only focuses on post-classification into a small number of topics (10-20), we consider the task of large-scale topic classification in the context of Twitter where the topic space is 10 times larger with potentially multiple topic associations per Tweet. We address the challenges above and propose a novel neural model, that (a) supports a large topic space of 300 topics (b) takes a holistic approach to tweet content modeling – leveraging multi-modal content, author context, and deeper semantic cues in the Tweet. Our method offers an effective way to classify Tweets into topics at scale by yielding superior performance to other approaches (a relative lift of 20% in median average precision score) and has been successfully deployed in production at Twitter.

Recent low-resource named-entity recognition (NER) work has shown impressive gains by leveraging a single multilingual model trained using distantly supervised data derived from cross-lingual knowledge bases. In this work, we investigate such approaches by leveraging Wikidata to build large-scale NER datasets of Tweets and propose two orthogonal improvements for low-resource NER in the Twitter social media domain: (1) leveraging domain-specific pre-training on Tweets; and (2) building a model for each language family rather than an all-in-one single multilingual model. For (1), we show that mBERT with Tweet pre-training outperforms the state-of-the-art multilingual transformer-based language model, LaBSE, by a relative increase of 34.6% in F1 when evaluated on Twitter data in a language-agnostic multilingual setting. For (2), we show that learning NER models for language families outperforms a single multilingual model by relative increases of 14.1%, 15.8% and 45.3% in F1 when utilizing mBERT, mBERT with Tweet pre-training and LaBSE, respectively. We conduct analyses and present examples for these observed improvements.

While large-scale pretrained language models have been shown to learn effective linguistic representations for many NLP tasks, there remain many real-world contextual aspects of language that current approaches do not capture. For instance, consider a cloze test “I enjoyed the _____ game this weekend”: the correct answer depends heavily on where the speaker is from, when the utterance occurred, and the speaker’s broader social milieu and preferences. Although language depends heavily on the geographical, temporal, and other social contexts of the speaker, these elements have not been incorporated into modern transformer-based language models. We propose a simple but effective approach to incorporate speaker social context into the learned representations of large-scale language models. Our method first learns dense representations of social contexts using graph representation learning algorithms and then primes language model pretraining with these social context representations. We evaluate our approach on geographically-sensitive language modeling tasks and show a substantial improvement (more than 100% relative lift on MRR) compared to baselines.

We evaluate a simple approach to improving zero-shot multilingual transfer of mBERT on social media corpus by adding a pretraining task called translation pair prediction (TPP), which predicts whether a pair of cross-lingual texts are a valid translation. Our approach assumes access to translations (exact or approximate) between source-target language pairs, where we fine-tune a model on source language task data and evaluate the model in the target language. In particular, we focus on language pairs where transfer learning is difficult for mBERT: those where source and target languages are different in script, vocabulary, and linguistic typology. We show improvements from TPP pretraining over mBERT alone in zero-shot transfer from English to Hindi, Arabic, and Japanese on two social media tasks: NER (a 37% average relative improvement in F1 across target languages) and sentiment classification (12% relative improvement in F1) on social media text, while also benchmarking on a non-social media task of Universal Dependency POS tagging (6.7% relative improvement in accuracy). Our results are promising given the lack of social media bitext corpus. Our code can be found at: https://github.com/twitter-research/multilingual-alignment-tpp.