In recent times, it is challenging to keep learners continually engaged and motivated in learning activities through traditional learning systems. Smart gadgets, computers, tablets, and electronic devices attract learners, often causing distraction and loss of focus from the desired objective and a disturbed learning process. Over the past few years, there has been a strong inclination to make use of these gadgets and devices to make teaching and learning process effective and fruitful. There is reported evidence where these devices, embedded with technological advancements both in hardware and software fronts have supported the learner tremendously. According to Conlan et al. (2002), Traditional learning systems transformed over time and have been existent in various forms such as Computer Based Training (CBT) systems, Web-Based Training (WBT) systems, e-learning systems, Multimedia learning systems and tools and Intelligent Tutoring Systems (ITS) to name a few significant ones. One of the intends has been to make learning interaction, more accessible, user-friendly, interactive, and responsive.

A revolutionary change of recent times has been the advent of machine learning and its integration in teaching-learning leading to the development of systems and processes, that to a reasonable extent can mimic a human instructor. One of the common interventions of ML has been learner monitoring through the algorithms over a large collection of statistical data available through review of assessment mechanism, such as tests, quizzes, and assignments made available to the learner (Mohri et al., 2013). Another one is self-determining structures from given learning information (learner response) and, in view of these structures, to create methodologies for teaching-learning.

A collector of procedures, where a teacher identifies learning needs, builds up the learning goals, plans instructional activities, executes the plan to meet intended learner outcomes, is termed as teaching and learning process (Keleş et al., 2009). It can be characterized as s process that has the potential to cause a permanent change in an individual (knowledge, ability, and attitude). It serves to be a means to achieve desired changes in a learner.

Teaching-Learning is one of the oldest phenomena, that has been existent from eternal times. It has been undergoing several transformations. While, the primary and essential goal has always been to make learning happen, the methodologies employed have been varying from time-to-time, that has led to different Teaching Learning systems.

The section below outlines, frequently observed and the types of teaching-learning systems.

E-learning is an instructive strategy which utilizes computers as teaching gadgets. Computers are used as a communication medium for learners to gather learning material, that is, computers have a tendency to supplement the traditional system of reading the subject matter representing as a medium for providing information. The innovation envelops aspects originating from different fields of study, such as software engineering, instruction methodology, and digital technology. Over a period, e-learning has also adorned different forms, which have been referred by different names in literature, such as educational hypermedia, CBT, e-instruction, e-tutoring systems, Web-Based Training (WBT), and CBT agents. Additionally, interaction amongst learners has been through message platforms, bulletin sheets, and discussion platforms. These systems, make use of the internet backbone for connecting the learning material with the learner.

Another teaching-learning system of modern time system is an Intelligent Tutoring System (ITS) (Keleş et al., 2009). These systems facilitate teaching-learning, through support to educators in creating and overseeing courses, for the learners, as per the learning preferences and points-of-view. One of the most fundamental principles of the ITS is the Borrowing principle which states that a learner borrows the needed information from worked examples and connects the new information with the prior knowledge (Hwang et al., 2017).

Intelligent hints provided by the tutoring system reduce the unproductive time and make the learning process more effective by scaffolding the content dynamically and also providing feedback and hints as per the need. Such type of system can be further improved by embedding it with an emotion detection mechanism in order to effectively understand the strength of the learner and accordingly adopt a suitable teaching style.

Machine Learning (ML) is a multidisciplinary field involving different fields, such as Artificial Intelligence (AI), Statistics, Information Theory, Psychology, and Neurobiology. According to Kotsiantis et al. (2007), the goal of ML is to solve real-time problems with the use of a model that provides a good data approximation.

Typically, the ML algorithms have been used to solve realtime problems by using clustering, regression, classification, ranking, and dimensionality.

On the basis of various factors, such as availability of training data, test data, and evaluation of teaching methods, the ML algorithms can be distinguished as supervised learning, unsupervised learning, semi-supervised learning, and reinforcement learning.

In current times, ML has been playing a crucial role in the teaching-learning process. More recently, it has become all the more significant than as people need to learn more to adapt to the society. Initially, the challenge of ML is to work on the learning abilities of individuals, further facilitating an effective teaching learning process.

While, wide work is available indicating specific interventions of ML, in varied teaching learning systems, the scope of the present paper is limited to, web-based learning and intelligent tutoring systems.

The literature review would be focused on various traditional teaching learning systems. This study is limited under the Machine Learning algorithm for various teaching learning techniques like Computer-Based Training (CBT), e-learning, Web-based Training (WBT), etc. In this study the types of Machine Learning techniques or algorithms likes supervised learning, semi supervised learning, and unsupervised machine learning algorithms were also discussed.

This has led to probing towards understanding the applicability of these systems, for different individuals, also encompassing individuals with special needs. Also, subsequent sections of this paper are an attempt to shed light on accessible e-learning, its impact on teaching learning, and ML supported assessment procedure, in the online learning environment.

This section is a narration of ML interventions in teaching learning systems, focused towards summarizing the significance of ML, as a growing inevitable ingredient of presently available teaching learning systems.

Chen et al. (2000) used ML algorithms, database management system, and decision tree techniques on elearning teaching system logs supporting techniques to effectively verify the learner.

Hlosta et al. (2017) presented a self-learning framework utilizing an ML algorithm to discover learner's at-risk in another course, with no past historical data. This research showed that XGBoost accomplished the best execution.

Liu and d'Aquin (2017) utilized a supervised ML algorithm to predict the performance of learners. They examined how statistic factors and internet learning exercises influence a learner's performance. Besides, they used various clustering algorithms to discover the group of slow learners who need extra assistance from the teacher. They presumed that the successful grouping of learners, yielded benefits and majority of these learners finish their higher education.

Mythili and Shanavas (2014) worked on predicting learning skills for learners with special needs. With reference to, their work, the learning skill depends on several general common characteristics like handwriting skill, speaking skill. Furthermore, the collected data over those attributes applied ML-based J48 and SVM classifiers. Amongst these classifiers, SVM showed maximum efficacy.

Mi-Children.org (2014) reveals the progress of research in the direction of understanding the learner levels using ML algorithms to facilitate the learner with more appropriate multimedia content, leading to learner enthusiasm and motivation. The idea is to provide suitable content through a small interface to make learning effective.

The expression "intelligent", when applied in teaching and learning context, refers to any behavior which if performed by a human educator would be considered as "good teaching". The ambition to develop, intelligent teaching machines dates back to the origin of the development of computers and has progressed further with the advent of ML field, where the endeavors in current time have been acknowledged through teaching machines in various degree and structures (McDermott, 1990).

ML is used for learning to act in a stochastic environment by interaction with the environment. When a learner interacts with an ITS, there is no perfect method for predicting the learner performance (response) at a given time, so designing an agent that will learn the strengths and weaknesses of the learner as they forge through each problem will assist in exposing helpful elements of the system that can then be exploited in order to make the learner effectively use ITS (Sutton & Barto, 1998).

Traditionally, most research on CBT learning has indicated advancement of techniques for solving teaching and learning problems and a huge number of the system, developed have been tried on simplified artificial issues (Reich, 1994). Subsequently, the research in the field of ML has generally been exceptionally rich covering hypothetical development. It, however, needs real-time applications with direct connection amongst hypothesis and practice. This is especially recognizable in the field of adaptive instructive hypermedia, where many developments are limited to few, hand-created systems, and omit thorough evaluation of utility.

Hollweck (2014) revealed an extensive range of meanings of methods, techniques, and procedures in teaching that are presented in manuscripts and aid in the philosophy of teaching (see, for instance, Richards & Rodgers, 1987). Be that as it may, the difference between all these is irrelevant for the present study (Pan et al., 2006). As the online training is not generally directed by a teacher, it depends mostly on the electronic feedback (electronic testing), that will justify the learning goals and time commitment of the learner towards the learning process.

Virtual amusement park system designed by researchers at the Institute of Software in the Chinese Academy of Science (Pan et al., 2006), utilizes graphics, audio, video, text, photos, and modern internet software to successfully build an e-learning system, which integrates the entertainment with learning. The learner can learn by playing in the virtual reality system (Gall & Breeze, 2008).

Recently, there has been growing interest and reported research concerning the utilization of ICT in teaching procedures. New advancements can be utilized as a part of teaching settings to improve learning, and numerous analysts have found that the use of ICT over a selected collection of school subjects can give an alternative, innovative, and creative process in the classroom (Gall & Breeze, 2008; Sutherland et al., 2004; Tacchi, 2005; Bjekić et al., 2014). ICT is a critical instrument for teaching effectively and teaching with creativity. The idea of constructivism and socio-constructivism is essential for various types of comprehensive teaching. ICT supported creatively and innovation further facilitates the development of products particularly targeted for disabled learners.

The use of ML, in products for the disabled, is widely available in the literature. Some that encourage reading ability are products with teachers, such as text-to-speech, speech to text, and spell checker.

Register et al. (2007) advantages for learners with disabilities as they undergo e-learning courses, tested as peer support by utilizing computer-based communication technologies providing possibilities for peer to peer collaboration and ways to stay away from social disengagement, providing support to learners with disabilities to be proactive and confident, as opposed to responsive and dependent, controllability of learning, The Adaptability in time and space managed by ICT modalities can address the special disabilities of learners. It allows learners to advance at their own particular pace and take responsibility for their own learning Multimodal correspondence or extensive variety of e-learning communication systems permit presentation of data in a way versatile to particular disability; individual learner educator communication can happen productively and effectively; asynchronous communication is an advantage for our learners with disabilities; and so on. Some authors have emphasized the viability of utilizing music as a therapy to enhance reading abilities of learners with disabilities (Haralick & Shanmugam, 1973). Considering, the study of reported literature, it is apparent that the ability to being in learner comfort, userfriendliness, compatibility, support, and adaptability, is a key to continued integration of ML in current and future learning systems and their variants.

Swathika and Sharmila (2017) proposed a way to deal with perceiving the emergency exit sign utilizing a cell phone. The researchers utilize edge detection and region detection techniques to distinguish possible object candidates in the picture captured.

The new methodology dependent on Support Vector Machines (SVM) has created promising outcomes. A SVM is a regulated learning calculation dependent on measurable relapse (Vapnik, 1998). The SVM calculation works by mapping the preparation set into a highdimensional space, isolating positive and negative examples.

This section elaborates the findings of the review. It has been observed that the web-based system using ML contribute more to the improvement of learner learning skills. The study indicated that learners with poor cognitive skills did not perform well, they looked for hints specific and failed to utilize general hints. On the other hand, a learner with higher abilities could perform better in a CBT system. They used online tools and requested hints from time to time as per their needs. Another finding summarized from this study is that the learners who need help were not being benefitted by the intelligent tutoring system since it offered help only if they requested. Therefore, it is recommended to design intelligent systems with a number of cognitive capabilities. Additionally, Intelligent hints provided by a cognitive model could reduce the unproductive time and make the learning process more effective. The findings of the study can be used in turn to enhance the design of online-learning systems. ML algorithms hold tremendous potential in terms of enriching the systems, facilitating the learning of individuals with special needs by providing versatility and adoptive learning experiences. These contribute to enabling them to build on their abilities and strengths, further enhancing their satisfaction, learning speed, and effectiveness.

A narrative review of teaching learning systems with respect to machine learning interventions has been undertaken and findings and recommendations have been discussed in this paper. Compiled findings are indicative of the link between learner abilities and their performance, when analyzed, considering the scenario, they undergo learning through specific learning systems. Further, enrichment of systems through ML has been directly co-related with improved learning effectiveness and this thought has been further extended to a recommendation for individuals with a disability, essentially with the deemed design alternatives.

This review work is performed at University of Petroleum and Energy Studies (UPES) with Project No SEED/TIDE/ 133/2016. The authors thankfully acknowledge the funding support received from Technology Interventions for Disabled and Elderly (TIDE) scheme under the Department of Science and Technology (DST). The authors show their gratitude towards the management of UPES for supporting research work.