Shi Ying

Turi Abu, Ying Shi, Thomas Fang Zheng et al.

We present a novel Automatic Speech Recognition (ASR) dataset for the Oromo language, a widely spoken language in Ethiopia and neighboring regions. The dataset was collected through a crowd-sourcing initiative, encompassing a diverse range of speakers and phonetic variations. It consists of 100 hours of real-world audio recordings paired with transcriptions, covering read speech in both clean and noisy environments. This dataset addresses the critical need for ASR resources for the Oromo language which is underrepresented. To show its applicability for the ASR task, we conducted experiments using the Conformer model, achieving a Word Error Rate (WER) of 15.32% with hybrid CTC and AED loss and WER of 18.74% with pure CTC loss. Additionally, fine-tuning the Whisper model resulted in a significantly improved WER of 10.82%. These results establish baselines for Oromo ASR, highlighting both the challenges and the potential for improving ASR performance in Oromo. The dataset is publicly available at https://github.com/turinaf/sagalee and we encourage its use for further research and development in Oromo speech processing.

Lantian Li, Dong Wang, Yixiang Chen et al.

Various informative factors mixed in speech signals, leading to great difficulty when decoding any of the factors. An intuitive idea is to factorize each speech frame into individual informative factors, though it turns out to be highly difficult. Recently, we found that speaker traits, which were assumed to be long-term distributional properties, are actually short-time patterns, and can be learned by a carefully designed deep neural network (DNN). This discovery motivated a cascade deep factorization (CDF) framework that will be presented in this paper. The proposed framework infers speech factors in a sequential way, where factors previously inferred are used as conditional variables when inferring other factors. We will show that this approach can effectively factorize speech signals, and using these factors, the original speech spectrum can be recovered with a high accuracy. This factorization and reconstruction approach provides potential values for many speech processing tasks, e.g., speaker recognition and emotion recognition, as will be demonstrated in the paper.

Xiangyang Jia, Carlo Ghezzi, Shi Ying

Constraint solution reuse is an effective approach to save the time of constraint solving in symbolic execution. Most of the existing reuse approaches are based on syntactic or semantic equivalence of constraints; e.g. the Green framework is able to reuse constraints which have different representations but are semantically equivalent, through canonizing constraints into syntactically equivalent normal forms. However, syntactic/semantic equivalence is not a necessary condition for reuse--some constraints are not syntactically or semantically equivalent, but their solutions still have potential for reuse. Existing approaches are unable to recognize and reuse such constraints. In this paper, we present GreenTrie, an extension to the Green framework, which supports constraint reuse based on the logical implication relations among constraints. GreenTrie provides a component, called L-Trie, which stores constraints and solutions into tries, indexed by an implication partial order graph of constraints. L-Trie is able to carry out logical reduction and logical subset and superset querying for given constraints, to check for reuse of previously solved constraints. We report the results of an experimental assessment of GreenTrie against the original Green framework, which shows that our extension achieves better reuse of constraint solving result and saves significant symbolic execution time.