Dr. Ayu Explains COVID-19 Social Media Monitoring and Chatbot MeTriase Expansion for Hospitals

BANDUNG, itb.ac.id – Humans as social creatures do one basic thing, namely to communicate verbally and nonverbally (in writing). The existence of the internet in our daily lives makes communication easier so that we can receive information or news quickly, especially during the COVID-19 pandemic that occurred in Indonesia since March 2020.

One technology that processes text or sound written in human language is also called Natural Language Processing (NLP). The Center for Artificial Intelligence ITB made a series 3 webinar entitled “AI-NLP for Communication during the Pandemic Period”, which was presented by a lecturer in the Informatics Engineering Study Program, School of Electrical and Informatics Engineering (STEI) ITB, Dr. Eng. Ayu Purwarianti, S.T., M.T., on June 5, 2020 at the ITB Artificial Intelligence YouTube channel.

Briefly, Ayu explained that in the NLP process there are inputs, processes and outputs. When humans speak, speech recognition will recognize the sound and then NLP will be processed. Similarly, documents, websites, social media. In the process, NLP has 3 main components namely natural language understanding (converting text or sound into a form understood by computers), natural language generation (issuing text in natural language such as chatbot responses), or text analytics (processing data with output in the form of dashboards or scoring).

“The benefits of NLP technology are improving user experience in terms of spelling checkers, word or text recommendations, being automate support (in the form of chatbots), as well as being the search engine and text translation that we often use,” Ayu explained.

In her website, Ayu discusses chatbot and media monitoring. Ayu explained that the chatbot development was done because the community was accustomed to using chat applications (WhatsApp, LINE, Telegram, and others). “The chatbot position is very strategic and has many advantages such as 24/7 availability, and is useful for agencies to spread information (promotive),” she said.

She explained, currently chatbot still has some weaknesses such as the ability to understand complex statements, need a large database of question-answer pairs, and less natural chatbot interaction. This study proposes several solutions to overcome these weaknesses.

For the first weakness, Ayu continued, the proposed solution is a hybrid approach, which is when chatbots can transfer questions automatically to human agents when chatbots realize that they have no knowledge of user questions. For the second weakness, the proposed solution is the “learning” mechanism of the human agent in the hybrid approach as well as the database input in the form of a document, not in the form of a question-answer list. As for the third weakness, the proposed solution is to add chatbot capabilities with speech recognition technology as well as the addition of the word normalization module to handle slang words.

The ITB Artificial Intelligence Center itself has developed Chatbot Frequently Asked Questions (FAQs) about COVID-19 by presenting true and valid information about the development of COVID-19 in Indonesia and the world. In addition, ITB collaborated with the Faculty of Medicine UI to develop the triage chatbot COVID-19. Chatbot is useful as a preliminary health check of patients in the hospital so as to minimize medical personnel exposed to COVID-19 from patients. Chatbot will classify patient emergencies into mild, moderate, and severe groups so that they can be handled directly according to their groups.

The second ITB research is COVID-19 Social Media Monitoring which can be accessed at https://covid19-socmed.id/. This platform aims to monitor the development of COVID-19 issues taken from social media (Twitter and Instagram) with analysis using related keywords, sentiments, and hoax classifications circulating in Indonesia. In the process, the NLP Engine on this platform is a text analytics that has three main components, namely concept extraction, sentiment classification, and hoax classification. The outputs from this platform are data presented in the form of a dashboard.
By using certain keywords, COVID-19 Social Media Monitoring can determine the trends of Indonesian society’s conversation. “We analyzed the use of keywords that are often used by Indonesian people before and when the Large-Scale Social Restrictions (PSBB) were enacted. After that, by using sentiment analysis, we can find out the topic has a negative, positive, or neutral response based on community responses. In addition, we can also find out the probability level of hoaxes on a topic, “said Ayu.
Another ITB research is the Hoax Classification. Hoax classification requires a hoax database and facts as initial capital for checking a text can be classified into hoaxes or facts. When compared with NLP for English, Ayu explained that the challenges in developing NLP for Indonesian are grammar and mixing language in one sentence. “Indonesians are accustomed to using more than one language in communication including one sentence. Besides that, in grammar we can understand the meaning of a statement written not in the SPOK order, “he said.
* The AI-NLP Webinar for Communication during the Pandemic Period can be watched through the following link: https://www.youtube.com/watch?v=rUTcTrzAVVw&t=1519s
Reporter: Billy Akbar Prabowo (Metallurgical Engineering, 2016). This news has previously been displayed on the ITB webpage.

Berita Terkait