Signal 5: Let an AI Robot Pick Up Your Phone Call on Civic Issues

Xiao Jing
Civic Analytics 2018
2 min readNov 9, 2018

Making a phone call to government department always wears out our patience in China. This is because we have to press extension numbers according to the recorded instructions before hearing a real recipient’s voice. Even then, people often spend a lot of time explaining requirements clearly to the other side since some elder people or people with lower education usually cannot speak in standard Mandarin and with dialects or local accents. Still they have less skills of the internet, not to mention via online channel. In that case, public engagement will get hurt.

By recognizing the logic form from the calls, the robot recipient can extract the meaningful word like verb and object in each sentences and use Seq2Seq model to predicate the following sentences, working as responses to the phone calls like human. In e-commercial business, the AI robot tech has been matured.

Incredibly, the “Equilid” a new model of natural language processing(NLP) technology is more than that. So called Socially-Equitable Language Identification1, Equilid improves the off-the-shelf AI solutions by “accounting for things like regional dialects, slangs or the natural way people talk when they regularly speak more than one language.” Said Stanford AI research & development team.

Comparing to traditional AI robot using Wall Street Journal or a German newspaper as phrase database2, the Equilid uses more additional phrase sources as European legislation and Wikipedia, Urban Dictionary, conversations about articles in the Talk section of the Wikipedia website.

If it is used in China office system, the database should add local TV talk show, and vlogs in local dialects from various provinces. With the help of AI robot, the workload of answering the phone calls will be greatly decreased and even the elder people who do not familiar with computers can use it. Still, we should slowly apply the robot from non-emergency department to crucial department and give citizens time to get used to it. However, talking with a real person will get people more comfort emotionally while using robots may let people feel the government cold.

1. Equilid: Socially-Equitable Language Identification.

https://github.com/davidjurgens/equilid

2. Stanford AI researchers make ‘socially inclusive’ NLP using Urban Dictionary and Twitter

https://venturebeat.com/2017/08/08/stanford-ai-researchers-make-socially-inclusive-nlp-using-urban-dictionary-and-twitter/

--

--

Xiao Jing
Civic Analytics 2018
0 Followers

Data Scientist@NYU CUSP, Journalist @Supchina