We know that artificial intelligence (AI) is smart enough to do a few things our minds cannot, and with incredible accuracy. And now, it seems it also has the capacity to detect loneliness in humans, which is an otherwise challenging task.
我們知道，人工智能足夠聰明，能以極高的精確度實(shí)現(xiàn)人類大腦做不到的一些事。而如今，人工智能似乎還能做到又一項(xiàng)極具挑戰(zhàn)的任務(wù) —— 檢測出人類的孤獨(dú)指數(shù)。
A new study, led by researchers at the University of California San Diego School of Medicine, US, has shown how AI tools can predict levels of loneliness from a person’s speech with an accuracy rate of 94 percent.
美國加利福尼亞大學(xué)圣迭戈分校研究人員發(fā)起了一項(xiàng)新研究，展現(xiàn)了人工智能工具如何通過一個人的語言來預(yù)測其孤獨(dú)程度，準(zhǔn)確率高達(dá)94%。
The study focused on 80 participants aged 66 to 94, a population particularly vulnerable to loneliness. The subjects were asked 20 questions from the University of California Los Angeles (UCLA) Loneliness Scale, which uses a four-point rating scale for questions such as “How often do you feel left out?” and “How often do you feel part of a group of friends?”
該研究聚焦66-94歲之間的80名測試者，這一人群尤其容易感到孤獨(dú)。加州大學(xué)洛杉磯分校的孤獨(dú)感量表準(zhǔn)備了20個問題，要求測試者作答，量表中使用了一個四點(diǎn)評分量表對一些問題進(jìn)行回答，例如：“你經(jīng)常感覺被冷落？”“你經(jīng)常感覺到自己是朋友中的一員嗎？”
They were also interviewed in private conversations, which were recorded and transcribed by researchers. The transcripts were then examined using natural language processing tools, including IBM Watson Natural Language Understanding (WNLU) software, to quantify expressed emotions.
測試者也接受了私人談話式的采訪，這些談話被研究人員錄音并記錄。記錄文本則使用包括IBM沃森自然語言理解軟件（WNLU）在內(nèi)的自然語言處理工具進(jìn)行檢驗(yàn)，從而量化這些表達(dá)的情緒。
The interesting thing about this system is that it not only uses dictionary-based methods, such as searching for specific words that express fear, but also presents corresponding patterns by testing the words used in the response.
這一系統(tǒng)的有趣之處在于其不光使用了基于詞典的方法，如檢索表達(dá)恐懼的特定詞匯，還能通過測試受試者回應(yīng)中的用詞體現(xiàn)相應(yīng)模式。
Varsha Badal, the first author of the study, noted that the WNLU software system uses deep learning to extract data from keywords, categories, emotions and grammar.
該研究的第一作者瓦爾沙·巴達(dá)爾稱，WNLU軟件系統(tǒng)使用深度學(xué)習(xí)，能從關(guān)鍵詞、類別、情緒、語法中提取數(shù)據(jù)。
“Natural language processing and machine learning can systematically examine long interviews from multiple individuals and explore how subtle speech features such as emotions may indicate loneliness,” Badal said. “Similar emotion analyses by humans would be open to bias, lack consistency, and require extensive training to standardize.”
“自然語言處理和機(jī)器學(xué)習(xí)能系統(tǒng)地檢查來自多位測試者的長時間訪談，并探索情感等微妙的語言特征是如何表達(dá)孤獨(dú)感的，”巴達(dá)爾表示?！叭祟愡M(jìn)行類似的情緒分析或許會存在偏見，缺乏一致性，并需要大量的標(biāo)準(zhǔn)化訓(xùn)練?！?/font>
The more lonely a person felt, the longer their responses to direct questions regarding loneliness. The system was capable of not just detecting the degree of loneliness in each subject, but also showing differences between the way men and women spoke about loneliness. The men were found to use more fearful and joyful words in their responses, while the women tended to acknowledge feeling lonely during interviews.
一個人越感到孤獨(dú)，對于有關(guān)孤獨(dú)的問題回答也越長。這一系統(tǒng)不光能檢測到每個話題中的孤獨(dú)程度，還能體現(xiàn)男女在談及孤獨(dú)時的不同表達(dá)方式。研究發(fā)現(xiàn)，男性會在回答中使用更多與恐懼、喜悅相關(guān)的詞匯，而女性則是會在采訪中承認(rèn)感到孤獨(dú)。
Co-author Dilip Jeste said that the IBM-UC San Diego Center is now exploring natural language patterns of loneliness and wisdom, which are inversely linked in older adults. “Speech data can be combined with our other assessments of cognition, mobility, sleep, physical activity and mental health to improve understanding of aging and to help contribute to successful aging,” he said.
聯(lián)合作者迪利普·杰斯特表示，IBM-加州大學(xué)圣迭戈分校中心正在探索孤獨(dú)和智慧的自然語言模式特征，這些特征在老年人群中呈現(xiàn)負(fù)關(guān)聯(lián)?！罢Z言數(shù)據(jù)能與我們對于認(rèn)知、運(yùn)動、睡眠、生理活動以及心理健康的其他評估相結(jié)合，從而增強(qiáng)我們對衰老的理解認(rèn)知，并有助于我們成功度過老年生活，”他說道。
(Translator & Editor: Wang Xingwei AND Luo Sitian)
http://www.cycjq.cn/story/3636.html
