THE USE OF SPEECH RECOGNITION TO FIND OUT READING ERRORS ON GOOGLE TRANSLATE

P


P.175
Translate can translate into 80 languages in the world. 5In Google Translate some features can make it easier for users to translate the language they want to translate, such as the write-your-own feature, image or camera feature, speech or voice feature, document feature, and others. 6The speech or voice feature of Google Translate is one implementation of the use of speech recognition or voice recognition.In the opinion of Andriana et al, speech recognition is a method carried out by a computer to recognize sounds made by humans without regard to the person's identity.The pronunciation time and vowel sound emphasis level which will later be aligned with the existing template in the database are references to be used as parameters 7 and speech recognition using the Google Speech API can recognize spoken sounds and then convert them into text. 8ccording to Amrizal and Aini, speech recognition has advantages and disadvantages.These advantages include, fast and easy to use.Meanwhile, the weaknesses in speech recognition are that it is prone to interference and the number of words known is limited. 9 Meanwhile, the definition of vocabulary, or what is called Mufrodat in Arabic according to Muchtar is a collection of selected words that will later create a language, and words are the smallest components of language that have independent characteristics. 10Language is a communication tool that plays an important role for humans to understand it, take advantage of it, and adapt it to various kinds of science and knowledge. 11Arabic is a language found in the Koran and also hadith.Arabic is a special language because Arabic is rich in meaning. 12In Indonesia, where the majority are Muslim, many people are interested in studying Arabic, with the aim of understanding and studying the content of the Koran and Hadith.Studying Arabic can be done at formal or non-formal schools 13 .Learning a language is not as easy as imagined, it takes time to learn the sciences related to Arabic itself. 14Among these sciences are language skills known as speaking skills, listening skills, writing skills, and reading skills. 15 learning Arabic, reading skills are among the main skills that must be mastered by Arabic language learners and are also an important factor in learning Arabic 16 .Reading skills are activities carried out based on the participation of several skills, namely scanning, digesting, and remembering. 17Reading skills are a communication process between readers and writers through written text. 18With reading skills, students will better understand and master everything, and students will also have more understanding and knowledge.Reading skills are the main initial capital for students.In reading skills, usually starting from the lowest level, starting with letters, then continuing to vocabulary, then continuing to sentences.In this research, vocabulary is the context that will be studied. 19According to Fitriliza and Fahmi, vocabulary

RESEARCH METHODS
This type of research is descriptive qualitative research with data collection techniques, observation techniques, and documentation techniques.The observation technique used by the researcher is a research observation table where during the research students read the vocabulary prepared in the table, then the researcher observes the results obtained and analyzes them and simple observations from previous journals related to speech recognition research.The data analysis technique that researchers used in this research is interactive data analysis.This interactive data analysis model was coined by Miles & Huberman.Interactive data analysis has three elements, namely; data reduction, data presentation, and conclusion.This research was carried out at SMP Muhammadiyah 35 Jakarta, taking a sample of 20 grade 7 students.

RESULT AND DISCUSSION
Based on research conducted by researchers using speech recognition on Google Translate to find out errors in reading 50 vocabularies on Google Translate, researchers observed the results of reading 50 Arabic vocabularies that researchers had prepared in the research instrument table.The following are the results of the research data that have been obtained, then the researcher counted the number of correct and incorrect numbers of correct letters and pronouns in each vocabulary seen by each student, and then obtained the following results:

P.182
Based on the results of the number of correct and incorrect counts calculated for each student, the researcher then converted the data into diagram form which aims to make it easier to understand, such as:

P.183
Based on the diagram above, the vocabulary gets the number of correct letters as 0, and the number of incorrect letters as many as 20.
Based on the diagram above, the vocabulary of gets a correct number of letters of 0, and several incorrect letters of 20.
Based on the diagram above, the vocabulary of gets the number of correct letters as 0, the number of incorrect letters as many as 20 Based on the diagram above, the vocabulary of gets a correct number of letters of 0, and several incorrect letters of 20.
Based on the diagram the vocabulary of gets a correct number of letters of 0, and several incorrect letters of 20.

P.184
The results obtained were based on research carried out by researchers regarding the use of speech recognition to identify errors in reading 50 Arabic vocabulary using Google Translate, so data was obtained as in the table and diagram above.In the diagram, several vocabularies have many errors in letter accuracy and some vocabularies have all letter accuracy correct.Some of the results obtained from using speech recognition from all vocabularies, only 1 vocabulary did not have several errors in letter accuracy, namely .
Then several vocabularies do not have the correct number of letters, namely The errors found in using speech recognition are very diverse, such as incorrect letter arrangements and the vocabulary being read does not match what is captured by the Based on the diagram above, the vocabulary gets a correct number of letters of 0, and several incorrect letters of 20.
Based on the diagram above, the vocabulary of gets a correct number of letters of 0, and several incorrect letters of 20.
Based on the diagram above, the vocabulary of gets 0 correct numbers of letters and 20 incorrect numbers of letters.
The Use Of Speech Recognition To Find Out Reading Errors On Google Translate Author: 1 Aulya Fitrie, 2 Ari Khairurrijal Fahmi   Submited: 2023-08-14  Revised: 2023-12-01  Accepted: 2023-12-21   P.185 speech recognition function of Google Translate.Reading errors that are often found in vocabulary are errors in vocabulary that end in the letter ta' marbutah ‫.)ة((‬In its use, according to Amrizal and Aini, speech recognition has a weakness that is a bit of a nuisance to its users, a weakness that is often found during research, namely that it is prone to Interference.Speech recognition is prone to interference because the sound signal system is based on frequencies or waves.When the information in a sound signal has a wave structure that is the same as the disturbing frequency structure, it will be more difficult to sort out the interference in the sound signal. 26During the research, the research environment was very noisy with lots of noise from the students' voices, so the speech recognition on Google Translate was hampered in recognizing the voices of students who were reading the vocabulary.Apart from that, speech recognition also has limitations on the number of words recognized, so this can also trigger errors in the results displayed by Google Translate.The limited number of words is caused by speech recognition running based on similar data held in the system database. 27

CONCLUSSION DAN SUGGESTION
Based on the results of research on the use of speech recognition to detect errors in reading 50 Arabic vocabulary words on Google Translate, the following conclusions can be drawn; 1) When using speech recognition in Google Translate, errors were found in the accuracy of letters and pronouns for each student.2) In the research results, there is one vocabulary whose letters and pronouns are all correct, namely the vocabulary ‫ِي‬ ‫ف‬ and 10 vocabularies that do not have the correct number of letters, namely 3) Weaknesses in speech recognition affect the research results, so that several vocabularies ending in the letters ta' marbutah ‫)ة(‬ do not have the correct number, both in terms of letter accuracy and phrasing accuracy.This is due to the limited words available in the speech database system recognition.
Based on the research that has been carried out, the researcher makes several suggestions, as follows, 1) In the next development of the speech recognition system developed by Google Translate, it is better to update the vocabulary in the database system.2) It is hoped that in the next research, we will examine why Arabic vocabulary ending in the letter ta' marbutah ‫)ة(‬ ends in the letter ha' ‫)ه(‬ when using the microphone feature on Google Translate.
is a collection of words known by 12 ‫ارهنشاة‬ And P.178 pronouncing as well as increasing motivation to learn English pronunciation.This research focuses on analyzing the use of speech recognition in Google Translate to find errors in reading 50 Arabic vocabulary words.This research aims to find out what errors there are in reading 50 Arabic vocabulary words using speech recognition on Google Translate.
The Use Of Speech Recognition To Find Out Reading Errors On Google Translate Vol. 11.No. 2 Desember 2023 The Use Of Speech Recognition To Find Out Reading Errors On Google TranslateAuthor: 1 Aulya Fitrie, 2 Ari Khairurrijal Fahmi

Vol. 11. No. 2 Desember 2023
The Use Of Speech Recognition To Find Out Reading Errors On Google Translate