Mining Arabic Agricultural Problems For Intelligent Guidance / Ahmed Abd El Hamed Mohamed Beleity

Supervisor: Mohamed Awny

Thesis (M.A.)—Nile University, Egypt, 2019 .

"Includes bibliographical references"

Contents:
Chapter 1: Introduction ............................................................................................... 08
1.1 Chapter Preface ..................................................................09
1.2 Background .........................................................................09
1.3 Thesis Outline and Organization ........................................10
Chapter 2: Text and Data Mining…………. .............................................................. 11
2.1 Background .........................................................................12
2.2 Case Study Of: Data Mining in Agriculture sector
and NLP in Arabic language ..............................................30
Chapter 3: Problem and significance .......................................................................... 36
3.1 The Problem ........................................................................37
3.2 The Objective of the Research ............................................37
3.3 The Significance of the Research .......................................37
3.4 The Methodology ................................................................38
Chapter 4: Research Work Implementation …………. .............................................. 39
4.1 The Ultimate Goal...............................................................40
4.2 System overview .................................................................41
4.3 Implementation of the Developed System ..........................42
4.4 Verification and Validation of the system results: ..............49
Chapter 5: Conclusion & future research …………................................................... 51
References ....................................

Abstract:
Due to the vital importance of the Agriculture section in Egypt and its huge workforce at different levels of knowledge and literacy the Egyptian Government has created in 1970 an entity “Agriculture Research Center”.
The Agriculture Research Center in Egypt developed a website to deal with farmers’ problems online. At present, farmers’ access to such information is extremely limited; their queries are received and manually processed by agricultural experts. Ironically, due to a large number of queries the processing gets significantly delayed. Moreover, due to the lack of a reliable centralized source of information there is usually duplication of effort by experts when dealing with similar queries, or common situations. There are too many duplicate unorganized questions and it is hard to find the right answer or the category where for the right answer for a raised questions belongs to. Also, the data has too many unfiltered special characters and a mix of slang and standard language which makes it a bit difficult for users to search for specific information without the help of agriculture experts.
The drawback of the old manual process is that farmers across Egypt post their questions and concerns to the website of the Agriculture Research Center resulting in quite a number of repeated and duplicated questions. Each time a question is posted, a technical expert answers that question whether it is a repetition or not. Also, the data has too many unfiltered special characters and a mix of slang and standard language which makes it a bit difficult for users to search for specific information. The current research is an attempt to bridge the information and communication gap between the field subject matter experts and farmers, using data from the Agricultural Research Center of Egypt. This research mines and delivers solutions from a textual database of over 4,000 cases of agricultural issues, enabling farmers to make faster and more efficient decisions perhaps without the need to contact the experts.
This research develops a system where farmers could use it by simply asking the question and getting the closest and most accurate answer. The system will also ensure that the answer provided is for the correct crop. The raised question is inserted in the document as a vector of words. Using the TF-IDF numerical representation of words and the Cosine Similarity algorithm the document lines are being vectored where each vectored line is compared with the rest of the document lines to identify the similarities, and to eliminate the repeated unnecessary words in the questioned issue while highlighting the most significant one. The system is developed to identify all the similar questions.
In this research, firstly, data is collected from the textual datasets in the Agriculture Research Center. Secondly, by using Natural Language Processing (NLP) the system automatically picks the most significant words to identify each raised question. Then, it picks the correct answer after filtering the repeated words. Lastly, the system ensures that the answer provided is for the correct crop.

Text in English, abstracts in English.