Mt applications are designed to surmount the language barrier. the goal of present english-sindhi mt system is to do optimal translations for day-to-day communications and small structures. this research would definitely open up the avenues for mt to expand its wings (to cater the limited domains like news, weather forecast, scientific articles etc.). this english-sindhi mt project targets the two major approaches of mt: rule-based and statistical based and its comparative analysis with each possible linguistic structure with reference to english-sindhi language pair. on the one hand, it is difficult to distill complexity and recursive nature of language through rules whereas on the other hand, large number of computational resources for the computationally less-resourced language (here, sindhi) require lot of linguistics expertise which is difficult to find in the countries like india and pakistan. sindhi with its 53,410,910 speakers in pakistan and around 5,820,485 speakers in india , has influences from a local version of spoken form of sanskrit and from balochi spoken in the adjacent province of baluchistan. sindhi has a relatively large inventory of both consonants and vowels compared to other indian languages. sindhi has 46 consonants and 16 vowels. before the standardisation of sindhi orthography, numerous forms of the devanagari and lunda scripts were used for trading. for literary and religious purposes, an arabo-persian alphabet known as ab-ul-hassan sindhi and gurumukhi were used. during british rule in the late 19th century, a persian alphabet was decreed standard over devanagari. in india, the devanagari script is also used to write sindhi. a modern version was introduced by the government of india in 1948; however, it did not gain full acceptance, so both the sindhi-arabic and devanagri scripts are used. in india a person may write a sindhi language in either script. rule-based mt: anglasindhi is a rule-based translation methodology which has been customized within the framework of anglabharti. the present system does the translation from english to sindhi has been developed with the assistance of nlp lab at c-dac, noida. the system envisages the use of human assistance to improve the quality of translation. parallel corpus for statistical mt: a parallel corpus is a collection of text, paired with translation into another language. the importance of corpus linguistics itself came into fore with the prevalence of e-text. with the advent of common use of computers, sindhi language also got some share and e-content gradually started growing. though unicode standards in indian languages have helped grow the content, however, there is no readily parallel training data available for english-sindhi language pair. as a consequence of this, sindhi which represents certain linguistic phenomenon such as pronominal suffixes on verbs, complex morphology and divergent word order constructions; has never been explored from a computational perspective. to initiate the process of including sindhi in the world of language technology, i have built 30k manually english-to-sindhi translated corpora for smt experiments.this 30k parallel corpora (english-to-sindhi) is not divided into domains because the aim of this research is to build general domain mt; therefore, any english source sentence would be part of the corpora (even a single word or phrase) but it is classified into different linguistic structures. there are number of precautions to collect source corpora to avoid noise and ambiguity. microsoft trsnaltor hub: english-sindhi mt: microsoft translator hub is an extension of the microsoft translator platform and service powered by windows azuretm. one can build a better-quality translation system easily, within a private website. the process of translation from one language to another can be broken into four steps: a) setting up a project; b) uploading data; c) training a system to translate from one language to the other; and d) deploying the system. this automated system i have trained on parallel document/texts (english and sindhi) which learns how words, phrases, sentences are commonly translated and how to process the appropriate context depending on the surrounding text. at the end, the invited reviewers will make recommendations on the results of training outputs to improve it further. this process can be repeated necessary until an acceptable level of quality has been achieved and the translation system is ready for deployment. after deployment of the system, the translation services are web-accessible using the standard microsoft translators apis-http, ajax, and soap interfaces as well as a web page widget. the use of microsoft translator hub is free. the present coverage of microsft translator hub for english-sindhi is 59.79%. moses english-sindhi mt: moses is an open source statistical machine translation toolkit which allows us to automatically train translation model (highest probability translation model) for any language pair. to train on this model, i have used same above mentioned 30k parallel document/text (english to sindhi). moses is licensed under the lgpl1. the primary development platform for moses is linux; however it can run on windows under cygwin. the installation of moses and supplement packages can be done on laptops with at least 2gb of ram and 10gb of free disk space and again depending on the size of corpora. the present coverage of moses english-sindhi mt is 46.56%.