e-ISSN: 2582-5208

International Research Journal of Modernization in Engineering Technology and Science

( Peer-Reviewed, Open Access, Fully Refereed International Journal )
Volume:05/Issue:12/December-2023 Impact Factor- 7.868 www.irjmets.com

HINGLISH TO ENGLISH TRANSLATION SYSTEM

Komal Potdar*1, Namrata Gaikwad*2, Meenakshi Sutar*3, Aditya Kurapati*4,
Ronak Dabade*5, Gaurang Khanderay*6
*1,2,3,4,5,6Department Of Multidisciplinary Engineering Vishwakarma Institute Of Technology,
Pune, 411037, Maharashtra, India.
ABSTRACT
The use of code-mixing, a linguistic phenomenon where speakers blend multiple languages, is widespread
among non-English speakers globally, particularly on social media. In the Indian context, individuals frequently
engage in code-mixing, combining English and Hindi in their online conversations, a practice colloquially
known as Hinglish. This linguistic fusion creates a vast amount of unstructured text on platforms such as social
media, blogs, and reviews. As code-mixing becomes increasingly prevalent, it poses a significant challenge to
machine translation systems. In this research paper, explore the algorithmic techniques developed to address
the complexities of handling code-mixed messages, specifically focusing on the Hinglish context. It discuss the
limitations of the system and its implications in the evolving field of text mining, shedding light on the modern
yet localized way of expression prevalent in Indian online communication. Through this study, the aim is to
contribute to the understanding of code-mixing challenges and advance the capabilities of natural language
processing systems in accommodating the diverse linguistic practices observed in social media discourse.
Keywords: Code-Mixing, Devnagari, English, Hinglish, Natural Language Processing, NLP, Transliteration,
Translation.
I. INTRODUCTION
In today's tech-driven world, it's becoming normal to mix different languages when people talk or post on social
media. People aren't using just one language as much anymore. This mix of languages makes a lot of data that
machines find tricky to understand. While there's been a bunch of work on translating pure languages, we now
need to pay more attention to studying and figuring out content in mixed languages. This study focuses on
translation model specifically for code-mixed language, like Hinglish, in Natural Language Processing (NLP).
This model helps machines better handle and make sense of the mix of languages when used in everyday
communication.
Hinglish is a language that mixes Hindi and English, blending them in conversations, individual sentences, and
even words. For example, you might say, "nahi mei nahi aa sakta," which means "no, I cannot come." This style
of speaking is becoming popular because it's a modern way of talking that still connects with local culture. In
India, a diverse country, there are people on both ends of the spectrum. On one side, there are Hindi speakers
who can read and understand the Devanagari script. On the other side, there are tourists from abroad who may
or may not fully understand the language. Since many local Indian markets, a big draw for foreign tourists, have
vendors who only understand Devanagari, there's a need to use Hinglish as a way for these two extremes to
communicate effectively.
The idea of this design is to restate Hinglish( Hindi English) which is combination of Hindi and English language
to pure English language. The proposed model operates by exercising Hinglish as a standalone language,
performing as a direct translator of code-mixed language into a pure form. This facilitates the analysis of
content expressed in mixed languages and serves to enhance the commerce between machines and humans,
fostering a more authentic communication experience.
II. RELATED WORK
S. H. Attri, T. V. Prasad, and G. Ramakrishna in [5] first determined if the sentence contained an expression or
idiom, then extracted it. The phrase was tokenized and classed as Hinglish, English, or Hindi, depending on its
original language. They then used morphological and reverse morphological analysis on each term. POS
Tagging sorted the words after analysis and Translation was carried out. "MujhE file send kar as soon as
possible" and "asap" were translated into Hindi. The resulting phrase in Hindi was "mujhE yathA shIghra
sanchikA bhEj" which translates to "Send me the file as soon as possible" in English. 12,000 Hinglish terms were

www.irjmets.com @International Research Journal of Modernization in Engineering, Technology and Science

[1354]
 e-ISSN: 2582-5208
International Research Journal of Modernization in Engineering Technology and Science
( Peer-Reviewed, Open Access, Fully Refereed International Journal )
Volume:05/Issue:12/December-2023 Impact Factor- 7.868 www.irjmets.com
labelled with idioms and translations. Pure Hindi sentences were much more accurate than pure English
sentences. Thus, Hinglish is Hindi with English words added, using Hindi syntactic and semantic components
instead of English ones. Because of this, it was found that Hinglish sentences translated into pure Hindi were
more accurate than those translate into English.
There is a lot of study being done on code-mixed material, especially on language tagging. The Jhamtani et al.
model was an ensemble model, (2014)[4] which was combination of two classi-fiers to form a LID mixed with
Hindi-English code. The first classifier employed features such as word frequency, modified edit distance, and
character n-grams, while the second classifier used the output from the prior one for the current word as well
as languages and pos tag for neighboring words to provide the final tag.
Authors in [2] propose a four-phase pipeline for automatic Hinglish-to-English translation. They also compare
"code switching" and "code mixing" and examine contextual problems such as "chalega" which meaning both
"moving" and "will It work?" This study used no comparable corpus. The language was tagged, transliterated
into Devanagari, translated from English to Hindi, combined with Hindi, and translated back into English.
III. METHODOLOGY
Overview
The methodology provides a systematic approach to develop Hinglish to English Translation System. It focuses
on accurate translation, replace short notation and user interactivity. It uses a dataset that contains Hindi
idioms for linguistic enrichment. It has different feature like virtual keyboard, virtual assistant, voice module
and file module.
Technological Stack
Frontend
The user interface is developed using HTML, CSS, and JavaScript, encompassing the design elements, virtual
keyboard functionality, and the integration of voice features.
Backend
Python is employed for backend development to handle translation logic, short notation replacement, and
interaction with the SQLite database.
Database
SQLite serves as the database, storing short notations, long notations, and a dataset of Hindi idioms.
Database and Dataset
The database is structured with two primary columns: "Short Notation" and "Long Notation". This design
facilitates efficient retrieval of long notations associated with replaced short notations.
The dataset consists of Hindi idioms in Devnagari script paired with their corresponding English translations.
This dataset enhances the linguistic component of the translation system, contributing to improved accuracy
and coverage.
Proposed System
The system translates the user's Hinglish text to pure English. Long notations take the place of short notations
in the text. It also provides the text box with a speech capability so that people may hear it. Finding and
replacing all brief notations in Hinglish text and translating it into full English are the primary objectives of the
current study. From a provided Hinglish text, the model recognizes the short notation and extracts it. Hindi
translations of the remaining texts will be provided. After that, the module receives the Hindi text and
translates it into English. We'll use the dataset to identify and replace all of the short notations with long ones.
The system has several input modes analogous as keyboard, voice, or in a train, as well as virtual keyboard as
an early- stage creations of the design. Input type file is useful because voice type will be useful if the user can't
write/ type properly or if the user chooses to submit the Hinglish text file. In addition, the virtual assistant is
assigned to answer the user's questions.

www.irjmets.com @International Research Journal of Modernization in Engineering, Technology and Science

[1355]
 e-ISSN: 2582-5208
International Research Journal of Modernization in Engineering Technology and Science
( Peer-Reviewed, Open Access, Fully Refereed International Journal )
Volume:05/Issue:12/December-2023 Impact Factor- 7.868 www.irjmets.com

Figure 1: Proposed Architecture

IV. RESULTS
As shown in the given figures, a text is passed to the system which contains a short notation and Hinglish text.
The system first checks for short notation in the given input. If any present then converts in the corresponding
long notation from the database. Then it checks for English words in the given input and find its Hindi meaning.
It performs transliteration and converts the whole text into devnagari script. Then system checks for idiom and
fetch data from the dataset. Then translation is performed and final output is given.

Figure 2: Output1
www.irjmets.com @International Research Journal of Modernization in Engineering, Technology and Science
[1356]
 e-ISSN: 2582-5208
International Research Journal of Modernization in Engineering Technology and Science
( Peer-Reviewed, Open Access, Fully Refereed International Journal )
Volume:05/Issue:12/December-2023 Impact Factor- 7.868 www.irjmets.com

Figure 3: Output2
For any translation to be successful, the system needs input in proper format without any spelling mistakes and
grammatical errors otherwise it may get confused and will provide abnormal output. Fig.1 and Fig.2 outputs
are from console where step by step explanation is given.
V. LIMITATIONS
 Hinglish is most frequently encountered on informal platforms. Since informal writing rarely adheres to
punctuation, correct spelling, and correct grammar, this adds an additional layer of complication to the
translations.
 Several words used in Hinglish are also found in English [2], in such cases it creates an obstacle for the
model to determine the source language and thus choose between transliterations and translations.
 There are no standard spellings in Hinglish; most users rely on the phonetics of the word to determine its
romanised spelling [1], thus resulting in a variety of words with the same meaning but different spellings.
For instance, “Nahi”, “Nai”, “Nhi” all mean “No” in English. This makes complication for the model.
 A large number of Hinglish terms have several meanings that can be determined from the sentence context
alone [1]. An example of this could be the term “chalega” which in some sentences means “will work” and in
some sentences means “walk”.
VI. FUTURE SCOPE
The identified limitations in the current Hinglish to English translation system pave the way for promising
future developments. Future work could focus on refining the system to better handle the informal nature of
Hinglish by incorporating advanced natural language processing (NLP) techniques that account for variations
in punctuation, spelling, and grammar. Additionally, exploring context-aware machine learning models may
contribute to improved disambiguation of meanings associated with Hinglish terms, addressing the challenge of
multiple interpretations. Introducing dynamic spell-checking algorithms that adapt to the phonetic variations
in Hinglish could enhance the accuracy of translations. Furthermore, the integration of sentiment analysis and
cultural context recognition could add another layer of sophistication to the translation process, ensuring that
nuances unique to Hinglish expressions are accurately captured. Overall, future advancements could involve a
holistic approach, combining linguistic analysis, machine learning, and cultural context understanding to create
a more robust and contextually aware Hinglish to English translation system.
VII. CONCLUSION
This paper addresses the complexities of translating Hinglish to English, recognizing challenges related to
informal expressions and diverse linguistic variations. Users are advised to input accurate spellings for optimal
results. While the system is robust, further advancements, such as dynamic spell-checking, remain avenues for

www.irjmets.com @International Research Journal of Modernization in Engineering, Technology and Science

[1357]
 e-ISSN: 2582-5208
International Research Journal of Modernization in Engineering Technology and Science
( Peer-Reviewed, Open Access, Fully Refereed International Journal )
Volume:05/Issue:12/December-2023 Impact Factor- 7.868 www.irjmets.com
exploration. This project lays the groundwork for future research, aiming to create a more precise and
culturally sensitive Hinglish to English translation experience.
VIII. REFERENCES
[1] Jadhav, I., Kanade, A., Waghmare, V., Chandok, S.S., Jarali, A.: Code-Mixed Hinglish to English Language
Translation Framework. In: 2022 International Conference on Sustainable Computing and Data
Communication Systems (ICSCDS). pp. 684–688. IEEE (2022).
https://doi.org/10.1109/ICSCDS53736.2022.9760834.
[2] Dr. S.V. Kedar, Sakshi Bhangale, Kunal Deokar, “Translation: Code-Mixed Language (Hinglish) to
English”, IJARSCT Volume 2, Issue 3, May 2022.
[3] IJCSI International Journal of Computer Science Issues, Vol. 11, Issue 5, No 2, September 2014 ISSN
(Print): 1694-0814 | ISSN (Online): 1694-0784www.IJCSI.org
[4] Attri, S.H., Prasad, T.V., Ramakrishna, G.: HiPHET: A Hybrid Approach to Translate Code Mixed
Language (Hinglish) to Pure Languages (Hindi and English). Computer Science. 21, (2020).
https://doi.org/10.7494/csci.2020.21.3.3624.
[5] Rao, Ashwini and DSouza, Nicole and Patel, Devarsh and Saravta, Jigyashu, Development & Study of
Hinglish to English Translation and Classification Techniques. Available at SSRN:
https://ssrn.com/abstract=4510958 or http://dx.doi.org/10.2139/ssrn.4510958.
[6] Chéragui, Mohamed Amine. Theoretical Overview of Machine Translation. Proceedings ICWIT.
2012.
[7] Hutchins W.J, Somers H L. An introduction to machine translation. London: Academic Press.1992:

www.irjmets.com @International Research Journal of Modernization in Engineering, Technology and Science

[1358]