Electrocorticographic (ECoG) spectral patterns obtained during language tasks from 12 epilepsy patients (age: 12-44 years) were analysed in order to identify and characterize cortical language areas. ECoG from 63 subdural electrodes (500 Hz/channel) chronically implanted over frontal, parietal and temporal lobes were examined. Two language tasks were performed. During the first language task, patients listened to a series of 50 words preceded by warning tones, and were asked to repeat each word. During a second memory task, subjects heard the 50 words from the first task randomly mixed with 50 new words and were asked to repeat the word only if it was a new word. Increases in ECoG gamma power (70-100 Hz) were observed in response to hearing tones (primary auditory cortex), hearing words (posterior temporal and parietal cortex) and repeating words (lateral frontal and anterior parietal cortex). These findings were compared to direct electrical stimulation and separate analysis of ECoG gamma changes during spontaneous inter-personal conversations. The results indicate that high-frequency ECoG reliably differentiates cortical areas associated with receptive and expressive speech processes for individual patients. Compared to listening to words, greater frontal lobe and decreased temporal lobe gamma activity was observed while speaking. The data support the concept of distributed functionally specific language modules interacting to serve receptive and expressive speech, with frontal lobe 'corollary discharges' suppressing low-level receptive cortical language areas in the temporal lobe during speaking.

The automatic identification of propaganda has gained significance in recent years due to technological and social changes in the way news is generated and consumed. That this task can be addressed effectively using BERT, a powerful new architecture which can be finetuned for text classification tasks, is not surprising. However, propaganda detection, like other tasks that deal with news documents and other forms of decontextualized social communication (e.g. sentiment analysis), inherently deals with data whose categories are simultaneously imbalanced and dissimilar. We show that BERT, while capable of handling imbalanced classes with no additional data augmentation, does not generalise well when the training and test data are sufficiently dissimilar (as is often the case with news sources, whose topics evolve over time). We show how to address this problem by providing a statistical measure of similarity between datasets and a method of incorporating cost-weighting into BERT when the training and test sets are dissimilar. We test these methods on the Propaganda Techniques Corpus (PTC) and achieve the second highest score on sentence-level propaganda classification.

The practices of high-frequency trading (HFT) are dependent on automated financial markets, especially those produced by securities exchanges electronically interconnected with competing exchanges. How did this infrastructural and organizational state of affairs come to be?Employing the conceptual distinction between fixed-role and switch-role markets, we analyse the discourse surrounding the design and eventual approval of the Securities and Exchange Commission's Regulation of Exchanges and Alternative Trading Systems (Reg ATS). We find that the disruption of the exchange industry at the hands of automated markets was produced through an interweaving of both technological and political change. This processual redefinition of the 'exchange', in addition, may provide a suggestive precedent for understanding contemporary regulatory crises generated by other digital marketplace platforms.

This paper brings together theories from sociolinguistics and linguistic anthropology to critically evaluate the so-called "language ideologies"-the set of beliefs and ways of speaking about language-in the practices of abusive language classification in modern machine learning-based NLP. This argument is made at both a conceptual and empirical level, as we review approaches to abusive language from different fields, and use two neural network methods to analyze three datasets developed for abusive language classification tasks (drawn from Wikipedia, Facebook, and Stack-Overflow). By evaluating and comparing these results, we argue for the importance of incorporating theories of pragmatics and metapragmatics into both the design of classification tasks as well as in ML architectures.