Articles

This paper describes the method that was used to produce additional acoustic model training data for the less-resourced languages of Lithuanian and Latvian. The method uses existing baseline speech recognition systems for Latvian and Lithuanian to align audio data from the Web with imprecise non-normalised transcripts. From 690 hours of …...

The run time of classical text-to-speech alignment algorithms tends to grow quadratically with the length of the input. This makes it difficult to apply them to very long speech recordings. In this paper, we describe and evaluate two algorithms that pre-segment long recordings into manageable “chunks”. The first algorithm is …...

We introduce a speech corpus containing multilingual code-switching compiled from South African soap operas. The corpus contains English, isiZulu, isiXhosa, Setswana and Sesotho speech, paired into four language-balanced subcorpora containing English-isiZulu, English-isiXhosa, English-Setswana and English-Sesotho. In total, the corpus contains 14.3 hours of annotated and segmented speech. The soap opera …...

The paper describes the process of creation of domain-specific speech corpora containing air traffic control (ATC) communication prompts. Since the ATC domain is highly specific both from the acoustic point-of-view (significant level of noise in the signal, non-native English accents of the speakers, non-standard pronunciation of some frequent words) and …...

Four open source formant trackers, three LPC-based and one based on Deep Learning, were evaluated on the same American English data set VTR-TIMIT. Test data were time-synchronized to avoid differences due to different unvoiced/voiced detection strategies. Default output values of trackers (e.g. producing 500Hz for the first formant, 1500Hz for …...

The paper describes an automatic Twitter sentiment lexicon creator and a lexicon-based sentiment analysis system. The lexicon creator is based on a Pointwise Mutual Information approach, utilizing 6.25 million automatically labeled tweets and 103 million unlabeled, with the created lexicon consisting of about 3 000 entries. In a comparison experiment, …...

In this paper, we propose a neural network model for Part-Of-Speech (POS) tagging of User-Generated Content (UGC) such as Twitter, Facebook and Web forums. The proposed model is end-to-end and uses both character and word level representations. Character level representations are learned during the training of the model through a …...

This paper describes the creation of the AIRBUS-ATC corpus, which is a real-life, French-accented speech corpus of Air Traffic Control (ATC) communications (message exchanged between pilots and controllers) intended to build a robust ATC speech recognition engine. The corpus is currently composed of 59 hours of transcribed English audio, along …...

Like many social variables, gender pervasively influences how people communicate with one another. However, prior computational work has largely focused on linguistic gender difference and communications about gender, rather than communications directed to people of that gender, in part due to lack of data. Here, we fill a critical need …...

This paper presents a multi-domain interests dataset to train and test Recommender Systems, and the methodology to create the dataset from Twitter messages in English and Italian. The English dataset includes an average of 90 preferences per user on music, books, movies, celebrities, sport, politics and much more, for about …...

The paper describes a recently-created Twitter corpus of about 6,000 tweets, annotated for hate speech against immigrants, and developed to be a reference dataset for an automatic system of hate speech monitoring. The annotation scheme was therefore specifically designed to account for the multiplicity of factors that can contribute to …...

With the enormous growth of Internet, more users have engaged in health communities such as medical forums to gather health-related information, to share experiences about drugs, treatments, diagnosis or to interact with other users with similar condition in communities. Monitoring social media platforms has recently fascinated medical natural language processing …...

The first 100 days corpus is a curated corpus of the first 100 days of the United States of America’s President and the Senate. During the first 100 days, the political parties in the USA try to push their agendas for the upcoming year under the new President. As communication …...

In this paper, we assess the challenges for multi-domain, multi-lingual question answering, create necessary resources for benchmarking and develop a baseline model. We curate 500 articles in six different domains from the web. These articles form a comparable corpora of 250 English documents and 250 Hindi documents. From these comparable …...

We present BioRead, a new publicly available cloze-style biomedical machine reading comprehension (MRC) dataset with approxi- mately 16.4 million passage-question instances. BioRead was constructed in the same way as the widely used Children’s Book Test and its extension BookTest, but using biomedical journal articles and employing MetaMap to identify UMLS …...

Recently, various datasets for question answering (QA) research have been released, such as SQuAD, Marco, WikiQA, MCTest, and SearchQA. However, such existing training resources for these task mostly support only English. In contrast, we study semi-automated creation of the Korean Question Answering Dataset (K-QuAD), by using automatically translated SQuAD and …...

Image question answering (IQA) is one of the tasks that need rich resources, i.e. supervised data, to achieve optimal performance. However, because IQA is a challenging task that handles complex input and output information, the cost of naive manual annotation can be prohibitively expensive. On the other hand, it is …...

The contribution of this paper is twofold: 1) we provide a public corpus for Human-Agent Interaction (where the agent is controlled by a Wizard of Oz) and 2) we show a study on verbal alignment in Human-Agent Interaction, to exemplify the corpus’ use. In our recordings for the Human-Agent Interaction …...

This paper reports an annotation to a corpus of database search dialogues on real estate, and the analysis on implicit information in the utterances for constructing database queries. Two annotators annotated 50 dialogues with a set of database field tags, resulting in a high inter-annotator agreement (Cohen’s κ=0.79), and the …...

Developing methods of automated inference that are able to provide users with compelling human-readable justifications for why the answer to a question is correct is critical for domains such as science and medicine, where user trust and detecting costly errors are limiting factors to adoption. One of the central barriers …...