Search

Automatic classification of disordered speech can provide an objective tool for identifying the presence and severity of speech impairment. Classification approaches can also help identify hard-to-recognize speech samples to teach ASR systems about the variable manifestations of impaired speech. Here, we develop and compare different deep learning techniques to classify the intelligibility of disordered speech on selected phrases. We collected samples from a diverse set of 661 speakers with a variety of self-reported disorders speaking 29 words or phrases, which were rated by speech-language pathologists for their overall intelligibility using a five-point Likert scale. We then evaluated classifiers developed using 3 approaches: (1) a convolutional neural network (CNN) trained for the task, (2) classifiers trained on non-semantic speech representations from CNNs that used an unsupervised objective [1], and (3) classifiers trained on the acoustic (encoder) embeddings from an ASR system trained on typical speech [2]. We found that the ASR encoder's embeddings considerably outperform the other two on detecting and classifying disordered speech. Further analysis shows that the ASR embeddings cluster speech by the spoken phrase, while the non-semantic embeddings cluster speech by speaker. Also, longer phrases are more indicative of intelligibility deficits than single words., Comment: Accepted at INTERSPEECH 2021

This work reports on the ongoing development of a multilingual pipeline for the summarization of European case law. We apply the TextRank algorithm on concepts of the EuroVoc thesaurus in order to extract summarizing keywords and sentences. In a first case study, we demonstrate the feasibility and usefulness of the presented approach for five different languages and 18 document sources.

In lots of natural language processing tasks, the classes to be dealt with often occur heavily imbalanced in the underlying data set and classifiers trained on such skewed data tend to exhibit poor performance for low-frequency classes. We introduce and compare different approaches to reduce class imbalance by design within the context of active learning (AL). Our goal is to compile more balanced data sets up front during annotation time when AL is used as a strategy to acquire training material. We situate our approach in the context of named entity recognition. Our experiments reveal that we can indeed reduce class imbalance and increase the performance of classifiers on minority classes while preserving a good overall performance in terms of macro F-score.

As supervised machine learning methods for addressing tasks in natural language processing (NLP) prove increasingly viable, the focus of attention is naturally shifted towards the creation of training data. The manual annotation of corpora is a tedious and time consuming process. To obtain high-quality annotated data constitutes a bottleneck in machine learning for NLP today. Active learning is one way of easing the burden of annotation. This paper presents a first probe into the NLP research community concerning the nature of the annotation projects undertaken in general, and the use of active learning as annotation support in particular.

Motivation: The recognition and normalization of textual mentions of gene and protein names is both particularly important and challenging. Its importance lies in the fact that they constitute the crucial conceptual entities in biomedicine. Their recognition and normalization remains a challenging task because of widespread gene name ambiguities within species, across species, with common English words and with medical sublanguage terms. Results: We present GeNo, a highly competitive system for gene name normalization, which obtains an F-measure performance of 86.4% (precision: 87.8%, recall: 85.0%) on the BioCreAtIvE-II test set, thus being on a par with the best system on that task. Our system tackles the complex gene normalization problem by employing a carefully crafted suite of symbolic and statistical methods, and by fully relying on publicly available software and data resources, including extensive background knowledge based on semantic profiling. A major goal of our work is to present GeNo's architecture in a lucid and perspicuous way to pave the way to full reproducibility of our results. Availability: GeNo, including its underlying resources, will be available from www.julielab.de. It is also currently deployed in the Semedico search engine at www.semedico.org. Contact: joachim.wermter@uni-jena.de

We report on the re-annotation of selected types of named entities from the Muc7 corpus where our focus lies on recording the time it takes to annotate these entities given two basic annotation units -- sentences vs. complex noun phrases. Such information may be helpful to lay the empirical foundations for the development of cost measures for annotation processes based on the investment in time for decision-making per entity mention.

Active learning is an effective method for creating training sets cheaply, but it is a biased sampling process and fails to explore large regions of the instance space in many applications. This can result in a missed cluster effect, which signficantly lowers recall and slows down learning for infrequent classes. We show that missed clusters can be avoided in sequence classification tasks by using sentences as natural multi-instance units for labeling. Co-selection of other tokens within sentences provides an implicit exploratory component since we found for the task of named entity recognition on two corpora that entity classes co-occur with sufficient frequency within sentences.