All Datasets

Product Features: AIGC-generated portrait data covering four styles—3D cartoon, comic, watercolor, and sketch—with each style including four ethnicities. The original portrait images are shared across styles, with about 10 images generated per person for each style. Ethnicities Collected: Black, White, Asian (non-Chinese), and Brown. Age Range: All adults, with a balanced gender ratio. Image Specifications: Resolution of 1080P or higher.

This Albania Albanian Pronunciation Lexicon, curated by DataoceanAI Inc., offers a wealth of linguistic resources tailored specifically for the Albanian language as spoken in Albania. With 53,542 meticulously crafted entries and an impressive 95.00% entry accuracy rate, this lexicon provides accurate pronunciation transcription in the popular XSAMPA phonemic system. It serves as indispensable training data for speech recognition, speech synthesis, and other language processing applications.

【Corpus Type】 Family, health, travel, education, work, gourmet food, marriage, movies, music, socializing, celebrities, weather, sports, and other common topics of daily life. Natural context, applicable to all industries. 【Pronunciation Person Information】 Gender: Male 45%, Female 55% Age: The pronunciation people mainly cover the age range of 16-45. Accent: Speakers are from Tirana.

This dataset was recorded in a quiet office/home environment, with a total of 318 speakers participating, including 179 males and 139 females. All speakers who took part in the recording were carefully selected through a professional screening process to ensure standardized pronunciation and clear articulation. The recorded texts encompass information from storybooks, textbooks, fairy tales, and other related content.

In this report, we focus on the recognition accuracy of the ASR systems, which is the most critical and commonly used parameter for measuring an ASRperformance.There are eight sections to the report. Section one is brief informaiton of the report. In section two, we briefly introduce the ASR systems and their models we use. In section three, we describe the data we prepare for this experiment. Next, in section four, we illustrate the methodology and evaluation metrics used for the accuracy of thesystems. And in section five and six, we present the results of the six ASR systems in detail from various dimensions. In the seventh section, we analyze the changes in these ASR systems during the last two evaluations. Finally, the eighth section concludes this report with further discussion and conclusions.

In this report, we will evaluate the performance of the 6 most representative PA engines, focusing on the result of accuracy, fluency, and prosody in utterance level, accuracy in word level, and accuracy in phoneme level, which are critical and commonly recognized parameters for measuring. For the following content of this report, we will first introduce the 6 engines evaluated and the features we use. Later, we will describe the data we prepared for this experiment. Next, we will illustrate the methodology and evaluation metrics used for the pronunciation assessment. Moreover, we will present the evaluation results and the analysis based on each aspect. Last but not least, this report ends with a conclusion and further discussion.

In this report, we focus on the recognition accuracy of the ASR systems, which is the most critical and commonly used parameter for measuring an ASRperformance.There are six sections to the report. Section one is brief informaiton of the report. In section two, we briefly introduce the ASR systems and their models we use. In section three, we describe the data we prepare for this experiment. Next, in section four, we illustrate the methodology and evaluation metrics used for the accuracy of thesystems. And in section five, we present the results of these ASR systems in sync mode from various dimensions. Finally, the sixth section concludes this report with further discussion and conclusions.

In this report, we focus on the recognition accuracy of the ASR systems, which is the most critical and commonly used parameter for measuring an ASRperformance.There are six sections to the report. Section one is brief informaiton of the report. In section two, we briefly introduce the ASR systems and their models we use. In section three, we describe the data we prepare for this experiment. Next, in section four, we illustrate the methodology and evaluation metrics used for the accuracy of thesystems. And in section five, we present the results of these ASR systems in async mode from various dimensions. Finally, the sixth section concludes this report with further discussion and conclusions.

This dataset was recorded in a quiet office/home environment, with a total of 114 speakers participating, including 59 males and 55 females. All speakers involved in the recording were professionally selected to ensure standard pronunciation and clear articulation. The recorded text covers information on animals, food, movies, etc.