Because of its flexible delivery model, cloud computing is increasingly important. A wide variety of tools are required to assist businesses with the ever quicker digital transition, from software (SaaS) systems (PaaS) to IT (IaaS). A new business segment is Machine Learning as a Service (MLaaS), giving businesses a simple way to process results. Platform-based service provides an operated ML platform, which makes it easier and quickest for developers to train and deploy their models. You will find a set of various frames and algorithms. The isolation of the framework of the application ensures that the own equipment is not handled additionally. Moreover, the scalability and elasticity benefit of a cloud-based network is also provided. The most talented resources are in model creation and checking using numerous algorithms. These four providers offer their applications with a command-line GUI. Google and Amazon also support REST APIs.
The Jupyter notebook also enables AWS, Azure, and IBM to run. Although Azure and IBM have only SDK for Python, several other programming languages are provided by Google and AWS. Both providers offer solutions to most popular ML problems (regression, classification) with their frameworks and algorithms, and in this regard, they cover the same set of functions. Specific secondary variables, including latency, accessibility of the system, or similar, may be present. In decision-making, play an essential role. Those, though, focus on the network, not on the interface. Besides the services offered by the platform, many ML APIs have been pre-trained with pre-trained models. For usage, no prior ML information is required; you send and receive your data from or from SDKs. Because of the already well-trained and validated models, broad models are not necessary to achieve performance. Typically, a REST GUI or SDK interfaces are made. According to the trained and tested models, a lot of research is not required, mainly since there is no need for separate training and test data sets.
III. Literature Survey
IV. Industry Technlogies
Natural Language Processing (NLP), which allows computers to communicate with human language, is the primary field for language and text processing. NLP requires, among other items, voice understanding, syntax analysis and text creation. In recent years, all these fields have made significant success not least because of profound learning developments (Young et al., 2018). In certain situations, language understanding, along with NLP, is used to recognize what users want. Virtual assistants such as Apple's Siri or Amazon's Alexa are the best-known manifestations of this development. The computer image recognition has made considerable strides over the last couple of years. Objects may be identified before the camera and classified with smartphones in real-time (Redmon & Farhadi, 2017). The coevolutionary neural networks (CNNs), which play especially an essential role in the identification of the face, can be credited to much of this performance (Matsugu et al., 2003). CNNs have achieved the lowest error levels, where all established picture detection algorithms are contrasted (Dan Claudiu Ciresan et al., 2011). Cloud-based applications offer a readily usable platform for incorporating pre-trained network output with your own devices. Face and facial detection are the primary focus of image processing services. These four vendors serve these roles. If looks are identified, all providers (bounding box) can determine the relative position in the image too. There are other ML APIs that are intended for specific projects, in addition to the image and text analyses. Azure Bot Service, Amazon Lex and Watson Assistant make it easier to build bots for users. Apps can combine search functions such as image search and video search and auto-completion with Azure Search APIs. Google promises a smoother career hunt and optimizes the recruiting cycle for businesses utilizing Cloud Talent Tool.
V. Platform Implementations
Amazon offers four different APIs for all language and text-relevant tasks, each with a specific area of responsibility. Amazon Transcribe is used for speech recognition. The audio files generated in the process can then be used for further analysis. An obvious use case would be the evaluation of customer calls. Transcribe supports English and Spanish. Amazon Polly offers speech synthesis. An existing text is reproduced with a computer-generated voice that is as human-like as possible. This is particularly useful for reading devices for the visually impaired. Polly supports 27 different languages and offers several voice variants for some expressions. For text analysis, become amazon Comprehend offered. Comprehend is an NLP service and offers various analysis tools to find new insights and connections in texts. Google also provides a speech recognition service with Cloud Speech-to-Text. A total of 119 languages and dialects are supported.
Cloud Text-to-Speech is available for speech synthesis. This service supports 56 different languages / dialects. These include individual WaveNet voices, which, according to Google, are perceived by people as more authentic than other speech synthesis technologies. WaveNet generates completely new audio files for the synthesis with the help of a neural network that is trained on realistic tone sequences and speech waves. This technology is also used in the Google Assistant and Google's translation service. Cloud Natural Language is the NLP counterpart to Amazon Comprehend and offers the same functionalities. With Cloud Translation, Google offers the possibility of integrating the public Google Translate translation service into your applications. With 104 supported languages, Google is ahead of the competition, although no statement on translation quality is included here.
With Cloud Translation, all words can also be translated among each other. Microsoft also offers language-relevant interfaces that form a sub-area of the so-called cognitive services, in which Microsoft combines its AI services. The Speech API includes services for text-to-speech (51 languages / dialects, sometimes several voices to choose from) or speech-to-text (30 words). The Translator Speech API combines these two services with a translation function and thus offers Speech-to-Speech or Speech-to-Text translations. Azure also provides a Speaker Recognition API that can be used to identify and verify people based on their voices. Among the Language, APIs you can find the analysis and translation tools from Azure. Text Analytics provides NLP for key phrase extraction, sentiment analysis, entity recognition, relationship analysis between words/phrases, and supports the identification of 120 languages. Tokenization is also possible via the Linguistic Analysis API. The Translator Text translation API supports 65 languages. Azure is also the only service to offer spelling and grammar checking via the Bing Spell Check API. IBM's Speech-to-Text Service supports nine languages , and the Text-to-Speech API can play eight different styles. The Language Translator Service offers the possibility to translate 21 words.
Natural Language Understanding analyses texts according to keywords, entities, relations, and moods and recognizes 62 languages. The service also supports other functions that are not available in the competition. Additional metadata can be extracted, or emotions related to the keywords and entities found can be queried. What also sets IBM apart from the game are Personality Insights and Tone Analyzer Services. With personality insights, more complex personality profiles can be created for the authors of the submitted texts. These profiles give, among other things, insight into character traits, possible needs, and tendencies towards consumer behaviour (e.g. preferred film genres). The Tone Analyzer examines the tone of speech (analytical, confident, reserved) and emotions of a text.
VI. Prototype Design
VII. Prototype Implementation
IX. Refernces List
X. Appendix A
XI. Appendix B