Jump to content
Chinese-forums.com
Learn Chinese in China

One Million Characters

  • entries
    17
  • comments
    75
  • views
    1,513

907,858 Characters to Go

murrayjames

355 views

Yesterday I finished reading a series of articles by IBM, in Chinese translation. The articles pertain to a commercial service called Personality Insights. The service helps companies find the right customers by analyzing text posted on Twitter and other social media platforms.
    
IBM claims it can predict human behavior by looking at Tweets. The Personality Insights service creates psychological profiles of Twitter users based on the words they use on the platform. According to the research that underlies Personality Insights, the words that people choose reveal aspects of their personality. By studying a person’s words, IBM can predict if that person is altruistic, trusting, orderly, self-disciplined, assertive, excitement-seeking, prone to worry or anger, artistic, intellectual, and more. The words people choose also reveal their needs and values. IBM claims it can predict if a person is likely to use credit cards when making a purchase, is susceptible to impulse buying, is likely to buy environmentally-friendly products or eat organic foods, is likely to start their own business, is likely to invest in a start-up, what kind of books and movies the person likes, etc.

 

Here is the sales pitch, in IBM’s own words:

 

Quote

通过Personality Insights 服务,企业可以将这一构想变成现实。通过利用企业和社交媒体数据的组合,企业可以应用分析来更深入地研究个人客户的特质。企业可以将这些结果用于更高质量的潜在客户开发、更精细的客户细分以及相关性和个性化都更高的专注市场营销。

 

The IBM Cloud website contains a demo of the Personality Insights service. I tried out the demo using some Tweets and other texts I wrote. I was unnerved but not surprised at how accurately the service predicted my personality. Maybe IBM predicted that, too.

 

Here’s a link to the demo:

https://personality-insights-demo.ng.bluemix.net

 

Since my last blog post I have also read three short works by 丁玲. The first work is《彭德怀速写》. It is a brief 1936 paean to celebrated CCP general Peng Dehuai. The second work is 《我怎样飞向了自由的天地》. It is a 1946 autobiographical essay. In the essay, 丁玲 writes on a number of topics: the May Fourth Movement, two feminist comrades that inspired her, her relationship with her mother, and her literary beginnings. The third work I read is 《夜》. It is a 1941 short story about a husband and a wife and their relationship in trying times.

 

The Personality Insights articles can be found on the IBM Cloud website. The website contains numerous articles; I read those describing and marketing the service, and explaining the personality research that undergirds it. Those articles are titled《关于》,《个性模型》,《大五类人格 - 随和性》,《大五类人格 - 尽责性》,《大五类人格 - 外倾性》,《大五类人格 - 情绪程度》,《大五类人格 - 开放性》,《需求》,《价值观》,《消费偏好》,《用例》,《使用指南》,《服务实战》,《服务背后的科学》, and《研究参考资料》. Here is a link:

https://cloud.ibm.com/docs/services/personality-insights/index.html?locale=zh-cn#about

 

Text of 《彭德怀速写》:
http://www.millionbook.com/mj/d/dingling/000/002.htm

 

Text of 《我怎样飞向了自由的天地》:
http://www.millionbook.com/mj/d/dingling/000/004.htm

 

Text of 《夜》:
http://www.millionbook.com/mj/d/dingling/dlzp/004.htm

 

Some statistics:
Characters read this year: 92,142
Characters left to read this year: 907,858
Percent of goal completed: 9.2%

 

List of things read:
《三八节有感》by 丁玲   (2,370 characters)
《我在霞村的时候》by 丁玲   (10,754 characters)
《在延安文艺座谈会上的讲话》by 毛泽东  (18,276 characters)
《自杀日记》by 丁玲   (4,567 characters)
《我没有自己的名字》by 余华   (8,416 characters)
《手》by 萧红   (7,477 characters)
《牛》by 沈从文   (8,097 characters) 
《彭德怀速写》by 丁玲   (693 characters)
《我怎样飞向了自由的天地》by 丁玲   (2,176 characters)
《IBM Cloud文档:Personality Insights》by IBM   (25,098 characters)
《夜》by 丁玲   (4,218 characters)



8 Comments


Recommended Comments

I was unnerved but not surprised at how accurately the service predicted my personality.

Out of curiousity I went to have a look at this, and they also will analyse from any text you have written and not just tweets, so I put in a bunch of my longer posts from here and let it do an analysis.

 

Personally, I think the results it predicts are like horoscopes - generic things that people then read in to, to see something that matches themselves.

Share this comment


Link to comment

I am a saxophonist and composer, and the demo predicted that I play music. That is a fairly specific thing to infer from a text that does not mention music at all.

My personality breakdown was a near-perfect match with the results of a Big 5 psychometric test I took a year ago. And my breakdown did not change much when I fed different texts, on different subjects, into the demo. Horoscopes do not work like that.

Share this comment


Link to comment
4 hours ago, murrayjames said:

My personality breakdown was a near-perfect match with the results of a Big 5 psychometric test I took a year ago

And mine was almost the complete opposite!

 

4 hours ago, murrayjames said:

And my breakdown did not change much when I fed different texts, on different subjects, into the demo

Now try putting someone elses texts in to the demo, and see how sentences in the description of your personality that you can still relate to.  If you still find things that are 'right', well, that is how Horoscopes work.

Share this comment


Link to comment
20 hours ago, imron said:

And mine was almost the complete opposite!

 

Hmm... interesting. I do see what you mean by horoscope sentences. One sentence that came up with someone else’s text—“You are motivated to seek out experiences that provide a strong feeling of belongingness”—is vague enough that people with very different personalities could relate to it.

 

The demo works for me, however. I put some texts written by other people into the demo. Many of the resulting sentences did not describe my personality accurately, though some did. The personality results using these texts were less relatable than the personality results using my own texts.

 

Also, when using my own texts, the demo spits out very similar Big 5 breakdowns, notwithstanding significant differences between the texts. So the demo is clearly tracking something in the way I write. Otherwise there would be more variation in these breakdowns, like there is when I use other people’s texts.

Share this comment


Link to comment

People definitely do have detectable writing styles (see the field of stylometry) so it's not a surprise that it can spit out similar breakdowns based on text written by the same person.

 

And I'm sure it does track something, my point of contention is how much that something has to do with your personality, and how much is it used to just choose out of several 'vague, but could be talking about me' sentences.

 

 

Share this comment


Link to comment

Yes, obviously people have writing styles. My point is that the demo is tracking something about the way I write AND what it infers from that something matches both the results of a Big 5 psychometry test and how I understand my own personality. When I tried using other people’s texts, many of the sentences were not “vague, but could be talking about me.” They did not describe my personality well at all.

  • Thanks 1

Share this comment


Link to comment
×
×
  • Create New...