lOMoARcPSD|55140700

Agriculture Chatbot Agribotusing Natural Language

Processing
computer engg (Dr. D.Y. Patil Institute of Engineering, Management and Research)

Scan to open on Studocu

Studocu is not sponsored or endorsed by any college or university

Downloaded by Mfatkhun Najaa (mfatkhun28@gmail.com)
 lOMoARcPSD|55140700

IEEE - 56998

AGRICULTURE CHATBOT(AGRIBOT)
USING NATURAL LANGUAGE PROCESSING
2023 14th International Conference on Computing Communication and Networking Technologies (ICCCNT) | 979-8-3503-3509-5/23/$31.00 ©2023 IEEE | DOI: 10.1109/ICCCNT56998.2023.10306512

Chenchupalli Chathurya Diksha Sachdeva Mamta Arora

Computer Science and Technology Computer Science and Technology Computer Science and Technology
Manav Rachna University Manav Rachna University Manav Rachna University
Faridabad, India Faridabad, India Faridabad, India
chathurya6302@gmail.com dikshasachdeva1202@gmail.com Imamta.arora@gmail.com

Abstract— Agri-bot is a text-to-text chatbot which

will help the farmers in answering the queries posed by II. OBJECTIVE
them, in text form. Through text-to-text communication, The research objectives are formulated based on the aim of
farmers can interact with the chatbot to obtain the study, which is as follows:
information on various aspects of agriculture, such as To develop a chatbot in different regional languages, mainly
weather updates, crop management practices, pest and
Hinglish and English.
disease control, and market prices for agricultural
products. The chatbot uses different Natural Processing ● To build the data of common queries and answers
Techniques like lemmatization and tokenization and it is related to agriculture.
trained using sequential Model. The text-to-text chatbot ● To use different NLP Techniques required in data
is made in English which shows the accuracy 80% and pre-processing.
Hinglish which shows the accuracy 85%. The use of an ● To experiment with different deep learning
agriculture chatbot can improve farmers' productivity, techniques to train the model.
reduce costs, and increase their access to information,
particularly in areas where internet access is limited. ● To evaluate the accuracy of the proposed model.

Keywords— CNN Models, Machine Learning, Deep III. LITERATURE SURVEY

learning, Natural Language Processing, Lemmatization, There are a lot of different agriculture chatbots available and
Tokenization. a lot of previous research work has been done in this field.
Here is some conclusive research work done that will help
I. INTRODUCTION
in building the Agri-bot:
Agriculture is the most important part of human life and
1. The study [1] has contributed a chatbot that helps
helps human life in one way or the other. Farmers play the
farmers in answering some agriculture queries that
most important role in the economy of the country. Farmers
might be faced by farmers. The dataset was collected
nowadays have evolved in so many ways whether it's related
from the agriculture website and some of the questions
to work or whether it is technology. The provision of smart
and answers were added which were in the XML file
technology in agriculture has not only improved the work
and queries are added that can be posed by farmers. The
but also made the work easier. Introducing different types of
model uses natural language which processes the query
technology like Climate Monitoring through IoT to farmers
and uses Recurrent Neural Network (RNN) algorithm to
has been of great help but it sometimes is a little bit tough to
convey the correct answer. Future work can be
understand. Thus, these technologies can be made easier to
contributed by adding more queries which can be easier
understand through a chatbot. The chatbot could be of great
for farmers.
help to the farmers, it would not only answer queries based
2. As it is demonstrated in [2], “Krishi Mitra” is an
on technology but also different policies that are regularly
agriculture chatbot that is in the form of virtual
introduced by the government. The Agri-bot is a text-to-text
conversation through which farmers can answer
chatbot that can help farmers by answering their questions.
questions related to agriculture. This chatbot also
The chatbot is made in two languages both in English and
informs all the facilities which are provided by the
Hinglish. Currently, WhatsApp and Facebook are one of the
government. The speech recognition after receiving
largest platforms where farmers are creating local
audio in the form of breaks or different pitches is fed
community groups to share their ideas and problems. So, we
into a neural network (RNN), which finds different
are creating a chatbot, called “AGRI-BOT”. A Bot that can
patterns from the different dialects and tries to respond
reply to the series of questions asked by the farmers
with a meaningful answer. The neural network
instantly in their regional language. Here the farmers can
performs such results by training data which makes it
ask agriculture-related questions like what type of soil to use
easier to predict the answers that the user may want.
or different government policies.

14th ICCCNT IEEE Conference

Authorized licensed use limited to: Mukesh Patel School of Technology & Engineering. Downloaded on March 21,2024 at 18:11:56 UTC from IEEE Xplore. Restrictions apply.
July 6-8, 2023
Downloaded by Mfatkhun Najaa
IIT - Delhi, (mfatkhun28@gmail.com)
Delhi, India
 lOMoARcPSD|55140700

IEEE - 56998

Future work is suggested to create a multilingual like whether, soil and fertilizer ailments. The service
conversational chatbot that can help some of the cannot provide 24*7 because the usage of the large
regional farmers to communicate with their regional crowd of farmers considered one of the answers is to
language. resolve the issue in AI via Bots. It can get admission to
3. A Smart chatbot as demonstrated in [3], helps farmers some of the websites, apps, social media, and so on.
in different ways related to different aspects of farming 7. Implementation of a Chatbot System using AI and NLP
like soil quality, weather, etc. A SASI (Smart [7]—It is very difficult to solve the query of students,
Agriculture System based), which is based on IoT is staff, or other persons. Before joining the college, the
developed, which monitors the condition of the field, student wants to know the college information, the
crop, minerals etc. with the help of built-in sensors. people do not know who is working in that college and
This system of Iota can help the farmers in reducing also if any queries the people are very confused to get
their workload. The data collected from the sensors are that information when the people opened the college
fed to the cloud-based database. This data is further website we can see the chat assistants automatically, it
used by the chatbot “Farmers Friend” through can appear the questions as well as solutions,
APIs.This Bot will further provide answers as asked by availability of just pressing the options. if the query is
the farmers with the help Of Natural Language resolved can ask the questions again. And also, people
Processing (NLP). The main advantage here is that the can know all the information easily without asking
bot can be hosted on messenger, web chat, or Twitter. other people. Here the interface is developed into two
The future work can be continued by making it into a parts one for the user and the other for the
conversational chatbot for the farmers who are unable administrator. All the information is stored in AIML
to read and write. files to store the question-and-answer pair. When the
4. As demonstrated in [4] a “Talk Bot” is a web input is matched the related answer will respond. The
application that takes voice notes as input from a farmer lemmatization can extract the keywords meaningfully.
in the form of a query that is related to farming and If any questions are similar to the files so that type of
provides a necessary answer for that particular query. combination will not be feasible. so, if the question is
The dataset used here is trained with hundreds of different, it can take the maximum threshold and give
questions and answers. After breaking voice note into the correct response.
minor words, the remaining text is pre-processed using 8. Analyzed the difficulty of medical aspirants [8] and
different NLTK libraries after which the Naive Bayes also who's endorsed by the sicknesses at different
Machine Learning algorithm is applied which detects conditions. AI can be used to predict the type of
for a similar question already present in the trained symptoms and solve people's problems by taking real-
dataset and thus answer accordingly. Future work can time data. If any of the affected people are tolerating
be done in creating a bilingual or a multilingual chatbot. any of the diseases, the patient knows the facts
5. A Farmer Bot is a chatbot [5] which interacts with the approximately the symptoms, which kind of meals the
farmers on a speech-to-speech basis. The dataset is patient needs to take, and the way they could save the
collected from KCC Dataset, which basically contains disorder. based totally on the users' comments they can
the log calls at KCC by farmers where each call broaden the queries as well as solutions. So here the
consists of eleven fields consisting of the crops, dataset was taken from the patient feedback and queries
pesticides, diseases etc. The farmer gives the input in in a different type of hospitals in India. They observed
the form of audio which with the the help of neural the 3 foremost analyses to realize the meaningful
network like LSTM, finds the patterns in the voice and content of the user. some of the customers do not
thus helps in predicting the future response. The future recognize the unique sentence that they ask so the state
work is to decrease the computing time by replacing of affairs of at least one of the phrases must be similar
LSTM with GRUs. and it gives the associated solution by way of using this
6. The maximum of the agriculture fields [6]is maintained analysis.
by uneducated villagers. India is likewise rapidly 9. The discussion is across the corona disease [9] and how
moving fast towards the generation. As we see a the farmers are affected due to the virus, so at that time
number of the farmers don’t recognize how to throw the chatbot become an initiative constructed which could
seeds consistent with the soil, so that time it is easy for collaboratively between the AI and NLP and it's far
them to free the crop. If any of the farmers get income occurred via the people in different regions. the dataset
from that crop so all farmers proceed with that approach became taken from the direct farmers around 2 hundred
and demand for that crop also increases however the individuals what are the issues they are going through,
uneducated farmers don’t realize the crop that isn't so the Agri-experts entered the problems inside the
suitable for that soil. The Agri-bot can solve the issues shape of queries. In this era every farmer has been

14th ICCCNT IEEE Conference

Authorized licensed use limited to: Mukesh Patel School of Technology & Engineering. Downloaded on March 21,2024 at 18:11:56 UTC from IEEE Xplore. Restrictions apply.
July 6-8, 2023
Downloaded by Mfatkhun Najaa
IIT - Delhi, (mfatkhun28@gmail.com)
Delhi, India
 lOMoARcPSD|55140700

IEEE - 56998

preserving telephones, so AI technology enforces the collected. Figure 1

chatbot that's helpful to the farmers like what crops they
want to domesticate, If the crop is struggling with any
of the diseases the farmers don’t know the sort of
medicines suitable to the type of sickness so chatbot
will genuinely helpful to those farmers via mentioning
the numbers, for uneducated farmers. In this paper, they
developed languages by way of the use of the 5
analyses in NLP which can be recognizing all sorts of
issues in that call records. right here the techniques they
followed amassing the records, for undesirable textual
content they used the pre-processing the information
technique and Tokenization. In the future, by
implementing chatbots in distinct languages and the use
of AI and NLP, the nearby farmers will fulfil this by Figure 1 .The Above diagram shows the Hinglish Dataset.
way of developing the nearby languages.
10. Artificial Intelligence performs [10]a chief role in any
The questions which were collected were day to day-based
sort of sector now, not only in technical-related fields,
questions that is questions, whose answers the farmers are
but it's also in agriculture fields. the dataset become
unaware of. The dataset was collected in English language.
taken from the Kisan call Centre [KCC]in India from
With the help of google translator the Hinglish dataset was
the year 2015 January to 2017, the information is
also made by translating English to Hindi but letters were in
available in the shape of recorded calls from Karnataka
English, thus Hinglish.
too. The real facts they took in a variety of around
80,012,856 calls, each call they could separate eleven
subjects which consists of records, time of calls, place,
crops (306 crop sorts), query, and KCC Agri-
professional furnished through the solution associated
with that question. In the dataset the facts are in textual
content layout before using the algorithms first off,
clean up the facts via the usage of Exploratory
information Analytics. After making use of the
tokenization Named Entity recognition and then
algorithms which can be getting the most accuracy
80%. In the future amass more datasets and grow the UI
that may interact with the farmers and get more
accuracy about 90%. Figure 2 . The above diagram shows the English Dataset.

IV. ABOUT THE DATASET V. METHODOLOGY

Agriculture dataset was collected using a questionnaire. The 1) DATA COLLECTION - The data in the chatbot was
motive of questionnaire was to collect agriculture related collected by taking the survey from different farmers.
questions and answers so that when any farmer wants to ask In total approximately eighty question and answers
questions related to agriculture the chatbot can answer it were collected. The data was collected in English, and
quickly thus reducing the time and effort of the farmers. then converted into Hinglish with the help of google
Approximately around hundred question and answers were translator. Hence two datasets were made one in
English and one in Hinglish. The dataset was stored in
two different JSON file.

14th ICCCNT IEEE Conference

Authorized licensed use limited to: Mukesh Patel School of Technology & Engineering. Downloaded on March 21,2024 at 18:11:56 UTC from IEEE Xplore. Restrictions apply.
July 6-8, 2023
Downloaded by Mfatkhun Najaa
IIT - Delhi, (mfatkhun28@gmail.com)
Delhi, India
 lOMoARcPSD|55140700

IEEE - 56998

Figure 3. The Above diagram the workflow of the

model
Figure 5 Accuracy and loss of Hinglish data

2) EDA TECHNIQUES USING NLP - From the Json 4) CHATBOT CREATION - A chatbot is made in code
file, the words were extracted and tokenized, which format itself which asks the query and predicts the
further were lemmatized. The words were also answers as close as possible.
converted to lowercase. The purpose of lemmatizing 5) Result - Figure 4 and Figure 5
words is to narrow everything down to the simplest shows the implementation and correct answers of
level it can be which thus reduces the time and error some of the questions which were asked in the
when it actually processes these words in machine chatbot.
learning.
3) DEEP LEARNING ALGORITHM- After
VI. RESULT
lemmatizing and tokenizing the words, sequential
model is applied on both the dataset i.e., English and The chatbot was made successfully in both English and
Hinglish. The sequential model is applied to train the Hinglish. The English chatbot showed accuracy around 80
model in predicting the answer of the query which is percent while Hinglish chatbot showed 89 percent accuracy.
asked by the farmer. The sequential model applied
consist of three layers. Then the program is trained
and the accuracy is obtained. The English chatbot
showed an accuracy of 76 % as shown in Figure 4
while Hinglish chatbot showed 85% as shown in
Figure 5.

Figure 6 Implementation of English dataset

Figure 4.Accuracy and loss of English data

14th ICCCNT IEEE Conference

Authorized licensed use limited to: Mukesh Patel School of Technology & Engineering. Downloaded on March 21,2024 at 18:11:56 UTC from IEEE Xplore. Restrictions apply.
July 6-8, 2023
Downloaded by Mfatkhun Najaa
IIT - Delhi, (mfatkhun28@gmail.com)
Delhi, India
 lOMoARcPSD|55140700

IEEE - 56998

[5] S. L. B, S. Naik, and P. M. Shreenidhi Student,

“‘Farmerbot’-An Interactive and Assistive Interface
for Farmers,” 2022, [Online]. Available:
www.ijcrt.org
[6] S. Gounder, M. Patil, V. Rokade, and N. More,
‘Agrobot : An agricultural advancement to enable
smart farm services using NLP’, 2021. [Online].
Available: www.jetir.org

[7] T. Lalwani, S. Bhalotia, A. Pal, S. Bisen, and V.

Rathod, ‘Implementation of a Chat Bot System
using AI and NLP’, International Journal of
Innovative Research in Computer Science &
Technology, vol. 6, no. 3, pp. 26–30, May 2018, doi:
10.21276/ijircst.2018.6.3.2.

[8] T. Lalwani, S. Bhalotia, A. Pal, S. Bisen, and V.

Figure 7. Implementation of Hinglish dataset Rathod, ‘Implementation of a Chat Bot System
using AI and NLP’, International Journal of
Innovative Research in Computer Science &
Technology, vol. 6, no. 3, pp. 26–30, May 2018, doi:
VII. CONCLUSION AND FUTURE WORK 10.21276/ijircst.2018.6.3.2.

In conclusion, an agriculture chatbot can be an important [9] S. Ayanouz, B. A. Abdelhakim, and M. Benhmed,
tool for farmers and those who are in the agriculture ‘A Smart Chatbot Architecture based NLP and
Machine Learning for Health Care Assistance’, in
industry. With the help of natural language processing and
ACM International Conference Proceeding Series,
machine learning , agriculture chatbots can advise and Mar. 2020. doi: 10.1145/3386723.3387897.
support farmers on a range of topics such as weather
conditions, pest control, soil management and issues related [10] K. Kumari, ‘AGROCHATBOT: Understanding the
to pricing. From the above results it can be concluded that Scope of Agriculture’, International Journal of
the Agribot chat bot is a successful chatbot which displayed Advanced Research in Science & Technology
the correct answers both in English and Hinglish. The (IJARST), vol. 7, no. 3, 2020, [Online]. Available:
www.ijarsct.co.in.
Future work can be done by creating the chatbot in speech-
to-speech format which would be easier to understand and
[11] Hasangi Kuhaduwa, Dilshan Pathirana, Pathum
operate. Liyana Arachchi, Vishma Dias, Surangika
Ranathunga, Upali Kohomban. “Question
REFERENCES Answering System for the Travel Domain”
Moratuwa Engineering Research Conference
[1] “Development of Agriculture Chatbot using (MERCon) pp 449-454 2017.
Machine Learning Techniques,” International
Journal of Innovative Technology and Exploring [12] Sharvari Gaikwad, Rohan Asodekar, Sunny Gadia,
Engineering, vol. 9, no. 2S, pp. 24–28, Dec. 2019, Vahida Z. Attar. “AGRI-QAS Question-Answering
doi: 10.35940/ijitee. b1008.1292s19. System for Agriculture Domain”. International
[2] U. Kshirsagar, J. Parte, M. Patil, and A. Aylani, Conference on Advances in Computing,
“Krushi Mitra: A Review of Agriculture Bots,” Communication and Informatics (ICACCI) pp
2020, doi: 10.26438/ijcse/v8i6.4550. 1474-1478 2017.
[3] V. P. Reddy S and N. K. Prasad S, “Farmer’s
Friend: Conversational AI BoT for Smart
Agriculture.” [Online]. Available: .
[4] Sanglap Sarka R, Venkatesh Rao, Baala Mithra S M,
Subrahmanya V R K Roa, “NLP algorithm Based
Question and Answering System”, IEEE 2015.

14th ICCCNT IEEE Conference

Authorized licensed use limited to: Mukesh Patel School of Technology & Engineering. Downloaded on March 21,2024 at 18:11:56 UTC from IEEE Xplore. Restrictions apply.
July 6-8, 2023
Downloaded by Mfatkhun Najaa
IIT - Delhi, (mfatkhun28@gmail.com)
Delhi, India