-Agricultural extension is the application of new knowledge and scientific research findings to agricultural practices through farmer education. As a result, agricultural extension agents or workers are people from government research institutes who educate or pass on information to farmers on how to use the new knowledge and scientific research findings. However, the conventional method of communicating the agricultural research outputs or findings to farmers through only face-to-face meeting has many challenges such as geographic dispersion between farmers and extension workers, poor communication capacity, poor transportation facilities, bad roads, inadequate funding and dialectical problems, which create great problems to effective communication of agricultural information to farmers. Furthermore, this mode cannot adequately handle urgent (time bound) information that should circulate within the farming populace. In this paper, a multilingual translation system was developed to enhance agricultural e-extension services delivery. The system employs a serial integration of rule-based and statistical machine translation techniques to translate agricultural information or scientific research findings from the extension workers in English (source language) into farmer's registered native or preferred language. Four (4) target languages were considered, which include Arabic, Hausa, Ibo and Yoruba. The system was implemented using Per Hypertext Processor (PHP) language version 5.3.5 and Structured Query Language (MySQL) version 5.0.2. The system integration test shows 65% accuracy in translating research outputs in English to farmer's registered native language. It is therefore recommended that the implemented system be adopted for its efficiency and accessibility in enhancing agricultural e-extension services delivery.
Extension agents or workers are people who possess an
acceptable level of edification, engaged by government to
bridge the fissure connecting the government and the farmers
in terms of agricultural services. These groups of people
educate or pass on information to farmers on how to use
information derived from government research institutes.
Such information empowers farmers to take control over
decision-making processes and resources for increased
productivity [
An effective agricultural extension depends on how fast and useful the extension services reach or meet the farmer’s information need. Thus, it is likely that the farmers benefit from agricultural research findings with the eventual goal of improving agricultural productivity.
Unfortunately, in most developing countries today,
extension agents still depend largely on traditional extension
approaches of transmitting agricultural information to
farmers. The traditional approaches of communication are
classified as one-way multipurpose and two-way
multipurpose sources. The one-way multipurpose sources
include: television, radio, public campaign, leaflet, pamphlet,
newspapers and magazines. While the two-way multipurpose
communication sources include: village fairs, field
demonstrations, trainings and study tours. Also included in
this category are: extension workers, private agencies,
paratechnicians, farm input dealers, non-governmental
organizations (NGOs), credit agencies, fellow progressive
farmers, output buyers/food processors and primary
cooperative societies [
These sources of transmitting agricultural information to
farmers are no longer effectual for urgent agricultural
research findings needed by farmers. However, the rising
face up to farmers in this new millennium is how to manage
with the information explosion and global trend in
agrotechnology. There is the need therefore, for inter-related
systems to diffuse information and technological innovations
to farming populace in developing countries [
In addition to the identified problems, farmers are even
more confronted with myriads of challenges now than ever
before in relation to information generation and
communication, which is as a result of lack of sufficient
agricultural knowledge and information that will enhance
farmer’s participation, collaboration, and integration in
agricultural decision making processes. Also, lack of
investment from other agricultural stakeholders, for example,
government agro-allied industries, and non-governmental
organizations have lead to low output in both production and
sustainability. It is also arguable that farmers’ illiteracy in
Information and Communication Technology (ICT) is so
high in the rural areas as most of the farmers lack access to
agricultural technology, innovations and education. There
exists also (i) dialectical disparity in which the research
findings are communicated, (ii) near absence of training and
re-training of both the extension workers and farmers on new
innovations [
Finally, the present understanding of Agricultural
extension is supporting people engaged in agricultural
production by facilitating, empowering and linking them to
markets and other players in the agricultural value chain; to
obtain information, skills and technologies to solve their
problems [
The remainder of this paper is organized as follows: section 2 presents the related work. Section 3 describes the research methodology used. The results and discussion of the system testing and evaluation were presented in section 4. Section 5 gives the concluding remarks.
II.
RELATED WORK
Machine Translation (MT) is one of the most essential
applications of computational linguistics that uses the
computer software or web application to translate text from
one language to another. One of the benefits of machine
translation is that it helps people to understand an unknown
language without the aid of a human translator. However,
MT is often perceived as low quality based on outdated
perception created by its use of older translation technologies
or freely available generic translation tools from Google or
Bing that have not been customized for a specific purpose
[
A few different types of machine translation are available in the market today. According to [9] the most widely used techniques include: Statistical Machine Translation (SMT), Rule-Based Machine Translation (RBMT), and Hybrid Machine Translation, which combine RBMT and SMT. These techniques are briefly explained as follows [9]:
The RBMT relies on countless built-in linguistic rules and millions of bilingual dictionaries for each language pair. The RBMT system parses text and creates a transitional representation from which the text in the target language is generated. This process requires extensive lexicons with morphological, syntactic, and semantic information, and large sets of rules. The software uses these complex rule sets and then transfers the grammatical structure of the source language into the target language. [9]. There are no human interventions during the conversion from one language to another language. Human intervention only takes place, if at all, after translation to manually correct errors in the machine translation output.
The SMT is a corpus based approach, where translation is generated on the basis of statistical models whose parameters are derived from the analysis of bilingual text corpora [9]. A massive parallel corpus is required for training the SMT systems. The SMT systems are built based on two probabilistic models: language model and translation model [9]. The merit of SMT system is that linguistic knowledge is not a requisite for building the system. The complexity in SMT system is creating massive parallel corpus.
HMT was built owing to the drawbacks of the two
approaches and their prospect to be integrated [9]. Statistical
and Rule-Based are two MT techniques, whose methods of
translation are orthogonal to one another. SMT do not need
to learn about the language at all, but RBMT is based on
gathering language rules. Due to this difference, integrating
or hybridizing SMT and RBMT gives a better performance.
The hybrid technique can be used in a number of different
ways. In some cases, translations are performed in the first
stage using a rule-based approach followed by adjusting or
correcting the output using statistical approach. In the other
way, rules are used to pre-process the input data as well as
post-process the statistical output of a statistical-based
translation. This technique is better than the previous two
and has more power, flexibility, and control in translation
[9],[
Kalna-Dubinyuk [
A mobile-based Agricultural Extension System was
developed by [
An implementation of e-extension system that uses
mobile phones for knowledge sharing was done by [
An ICT-based agricultural extension service delivery
system was proposed for Nigeria by [
E-sagu was implemented by [
Shrikant and Shinde [
E-agriculture framework was developed by [17]. The framework proposes an implementation of an e-farming system that can be used in aiding sustainable agricultural farming practices. The incorporation of IT into farming involves the integration of diverse technologies, with each capable of positively impacting the efficiency of farming activities, thereby, promoting sustainability in agricultural practices. This framework has overcome farmers’ literacy level problems since the framework proposed is meant to compliment and replace the traditional extension services delivery. The major limitation of this framework is that language translation from source language of the agricultural information to the target local language understood by the rural farmers is not taken in to account.
III.
A.
The proposed multilingual translation system is based on the hybrid machine translation technique, which is a serial integration of the rule-based and statistical machine translation techniques. It is designed to translate source text from English language into any of the four preferred target languages: Arabic, Hausa, Ibo and Yoruba. The operational process of the multilingual translation system, as shown in Fig 1, is divided into six modules: deforming and preediting, analysis, transfer, generation, reforming and postediting, and statistical error checking. These modules are explained as follows: 1) Start: This is the beginning of the process.
2) Input Text: When a source text/sentence (in English language) of Agricultural Information (AI) is entered as input, the following process ensued:
3) Deforming and Pre-Editing: This is a preprocessing module. In this module, deforming is performed as a process in which the machine checks the part of the source AI text/sentence that does not require translation such as pictures, figures, diagrams and identifies only the portion of the source text/sentence that can be translated. Similarly, pre-editing involves fixing up the punctuation marks that does not require translation. This is to make the machine language translation of the AI easier, faster and efficient.
4) Analysis: In this module, the source text of AI is analyzed based on the linguistic information provided to produce a complete parsing of a source language sentence. Thus, it comprises of two components: tagger and parser.
a) Tagger: This component identifies the linguistic property of individual word of AI in the source text through the following processes: (i) Morphological Analysis: This aspect determines the form of AI word such as number, tense or part of speech (POS) tagger. (ii) Syntactic Analysis: This determines whether
AI words are subject verb or object.
b) Parser: This component breaks AI words/sentence into smaller elements, according to a set of linguistic rules that describe its structure through semantic and contextual analysis which determines the proper interpretation of AI text/sentence from the result produced by syntactic analysis. This is achieved by using lexical and semantic analyzer created by parser.
5) Transfer: In this module, the syntactic/semantic structure of the AI source text is then moved in to the syntactic/semantic structure of the AI target languages.
6) Generation: In this module, lexical transfer (the mapping of a source-language lexical item with an equivalent target-language item) occurs and mapping dictionary entries into appropriate inflected forms to yield a target-language equivalent term. This is achieved using Arabic lexicon, Hausa lexicon, Ibo lexicon and Yoruba lexicon to ensure proper interpretation.
7) Reforming and Post-Editing: This is a postprocessing module. In this module, once the AI text is translated, the target text is reformed after post-editing. This involves re-incorporation of non-translated portion of the source AI to target text for quality and adequate target AI to be disseminated to famers.
accurate grammatical matching of the target output produced by the rule-based approach into its statistical approach equivalent. Translation error checking is done to ensure good quality translation of the target texts of the AI. But in a situation where the rule-based translation technique does not correspond with its statistical machine translation with the agricultural extension agent, receive/read agricultural information from their mobile phones and also send their comment/query/request to AEA. equivalent, a decision is taken to consider another set of linguistic rules starting through the analysis, transfer down to the generation stage using the rule-based technique until the target text matches that of statistical based target text.
9) Output Text: The output target texts/sentences are the proper and accurate equivalent translation of the source AI text (English) in to target AI texts (that is Arabic, Hausa, Ibo or Yoruba) meant to reach the farmers according to their registered target languages with the AEA.
10) Stop: This mark the end of the process.
The proposed framework for a multilingual translation system to enhance agricultural e-extension services delivery is shown in Fig. 2. The system connects three major stakeholders of an agricultural extension services namely the farmer, researcher/expert and agricultural extension agent. The role of each stakeholder is explained as follows: 1) Researcher/Expert: This stakeholder provides critical research output in response to specific needs of the farmers. The research output cut across different aspect of farming including crop farming, livestock farming and the rest. Whenever a research request get to the research institute, it is handled by an expert in the area requested who after his findings relays a feedback through extension agents to farmers. In this system, the researcher/expert login to the system, read information/research request and communicate research request and finding to the AEA. The medium of communication between the researcher and AEA is through the web and their language of communication is English.
2) Agricultural Extension Agent (AEA): This is a trained expert who serves as a link between the farmers, research institute and farm input firms. They convey information responsive to the requirement of any component in the system. The AEA usually pay regular visit to farmers, interact with them in order to know their problems and concerns and then send it to an expert requesting for a solution. The functions of the AEA in this system includes login, update of information, register and manage farmers, send updates of agro information via SMS to farmers and also send research request to researcher/expert. In addition, the AEA interact with researcher through the web-based system and send agro information to farmers from the system to their mobile phones (it could be ordinary phones or android phones).
3) Farmers: This is the consumer of agricultural information. This is the component that receives agricultural information from research institutes and farm input firms via agricultural extension agent relevant to their farming requirement(s). This is because timely delivery of agricultural information to farmers means empowerment to them. The multilingual translation system for enhancing agricultural e-extension services delivery is aimed to provide farmers with this timely information they required based on their registered local languages (Arabic, Hausa, Ibo or Yoruba). The role of the farmers includes: registering
Yes
Start Input text (that is
English) Deforming and Pre-editing
Analysis
Tagger - Morphological Analysis -Syntactic Analysis
Parser
Semantic and Contextual Analysis
Transfer: Re-Ordering Syntactic and
Semantic of source text
Generation
Syntactic and Semantic generation of target texts using target languages lexicons
Arabic lexicon Hausa lexicon
Ibo lexicon
Yoruba lexicon Reforming and post-editing
Arabic Corpus Hausa Corpus
Ibo Corpus Yoruba Corpus
Error No
Stop Output target
texts Statistical error checking using corpus to check for language translation conformity with the result obtained by rule-based approach Farmers · ·
·
S y s t e m E n v i r o n m e n t
Web Interface
A three (3) tier model was adopted for the proposed multilingual translation system which consists of front-end (presentation tier), middle tier and back-end (data tier) as explained below:
1) Presentation Tier: This is also known as the front end and at this level, information is presented to client (i.e. Researcher or AEA) via browsers. This tier was developed using Per Hypertext Processor (PHP) language version 5.3.5.
2) Middle Tier: This tier is also known as server side. It is used for processing request through the My Structure Query Language (MySQL). MySQL 5.0.2 was used. This is also where the language translation takes place.
3) Data Tier: This tier is also known as back-end. This is the data center for the system which uses the · · · · ·
MySQL database management software that store collection of information and organized them so that it can easily be accessed, managed, and updated. MySQL command is used to insert, update, fetch and delete data in this system database.
To ascertain the workability between unit functions of the implemented system (interoperability of function between components), system integration testing was carried out. Four (4) test cases were tested hundred (100) times each. The test result shows how each component responded to an event that identifies specific functions of the design to whether or not the responses are as expected.
Integration testing analysis is shown in Table I. Test case 01 shows that 80.3% SMS from AEA were sent and received while 19.7% were not. Test case 02 shows that 74.6% of the web messages from AEA to experts and from experts to AEA were successful while 25.4% were not.
On test case 03, 73% experts and extension agent’s login to the web application were authenticated by granting access to the web application functions while 27% was not. Test case 04 tested the accuracy of translation from source language (research output in English) to farmer’s registered native language, 65% translation was achieved to each of the target languages while 35% were not.
B.
The evaluation of the implemented system was done in order to validate what the research work proposes, and to have a thorough understanding of how well it is working.
In other to ascertain the effectiveness, efficiency and capability of the implemented system for enhancing agricultural e-extension services delivery using the multilingual translation technique, the implemented system was used as a pilot scheme with forty eight (48) respondents. That is thirty (30) farmers, ten (10) extension agents and eight (8) experts or researchers were selected within Suleja and Minna, Niger State, Nigeria.
Some questions were directed to the selected farmers, extension agents and experts and their responses were collected instantly. Simple percentage method (SPM) was used for the calculations as shown in Table II. Out of the forty eight (48) numbers of validation forms distributed only forty three (43) were filled and returned.
Results obtained from the evaluation of the usage of the developed system shows that respondents believed the system will bridge the information gap amongst researchers, extension agents and farmer functionalities, 41(95.3%) are satisfied that the linkage provided by the implemented system is adjudged by the respondents as the best ever, while 2(4.7%) thought otherwise. Meanwhile, 33(76.7%) are satisfied with usability and accessibility of the system for enhancing agricultural e-extension services delivery while 10(23.2%) were unsatisfied.
24(55.8%) were satisfied that the system provides outmost confidentiality on the researchers, extension agents and farmers information as well as security of the system. 19(44.2%) were not satisfied. Similarly, 32(74.4%) believe the implemented system is effective and efficient in enhancing agricultural e-extension services delivery while 11(15.6%) were not satisfied.
system has bridged the communication gap among researcher, extension agent and farmers? Are you satisfied with the user friendliness (easy to use) and accessibility of the system as adequate for enhancing agricultural e-extension services delivery? Are you satisfied with the level of security and confidentiality of farmers, extension agents and researchers information on the system?
efficiency and effectiveness of the system for enhancing agricultural e-extension services delivery?
Response (in number
and %) Yes/Satisfie
d 41 (95.3% )
Response (in number and
%) No/Unsatisfie
d 2 (4.7% ) 33 (76.7% )
10 (23.2%) 24(55.8% )
19 (44.2% ) 32 (74.4%)
11 (15.6%) 34 (79.06%) 9 (20.9%) On the general functionalities of the system 34(79.06%) are satisfied with the performance of the system functionalities, as such believe the implemented system is a veritable tool for enhancing agricultural eextension services delivery while 9(20.9%) were not.
CONCLUSION
The research work implemented a multilingual translation system for enhancing agricultural e-extension services delivery that ensures real time agricultural information is provided to farmers irrespective of their geographical location and language. The implemented system translates the agricultural information from a source language (English) into four(4) other native languages (Arabic, Hausa, Ibo and Yoruba) depending on which the native farmer reads and understand. The implemented system has also brought all the stakeholders (researcher, agricultural extension agent and farmers) in agricultural information generation and dissemination together by enabling the AEA to send farmers research request to researchers or experts and receive research findings from the researchers via the web-based application.
In addition, farmers receive instant text messages from the AEA via their mobile phones and on requests, queries made on agricultural information. The acceptance evaluation of the implemented system shows that the implemented system is efficient and effective for enhancing agricultural e-extension services delivery.