Quantifying Qualitative Data Big Time! Jukka Sihvonen University of Vaasa

Transcription

1 Quantifying Qualitative Data Big Time! Jukka Sihvonen University of Vaasa

2 Analysis workflow Numerical Data minute(s) minute(s) Data acquisition Data analysis Results Unstructured Data week(s) month(s) Quantitative - Replicable - Scalable - Easily revisable Web Service Qualitative - Not replicable - Not scalable - Revision is hard

3 Traditional way Unstructured data - Photos - Movies - Speech - Text Analysis at the office by hand Results

4 Application Programming Interface (API) Data refining API Data API Unstructured data - Photos - Movies - Speech - Text Analysis API Results Intelligence API Scalability Replicability Speed Possibilities

5 Data APIs retrieve machine readable data Twitter example: #GDPR Request... Response import twitter api = twitter.api(my_credentials) tweets = api.getsearch(term = GDPR, count = 10) Twitter API May :15 AM Waterford Chamber There's just over 10 days t... May :15 AM British Tinnitus A new law, the General Data Protection Regulat... May :15 AM trust-hub The 8 rights of individuals when it comes to #... May :15 AM Kamil Google offers an unusually clear view of how i... May :15 AM Achille Zappa New European #dataprotection rules... May :15 AM Tom Skeen Is your Data Protection Officer... May :15 AM Brian Rogers 90% of Companies Unprepared for GDPR Deadline... May :15 AM Paul Laughlin Are you having to turn back the #Marketing #An... May :15 AM zkipster #GDPR wipes out the commoditized third-party d... May :15 AM Cara Cardona Ahead of GDPR, Google has made it easier for u... Example APIs: - Wikipedia - LinkedIn - Facebook - Twitter - Instagram - Yle - PRH - Finlex - Finnish Parliament - EU Law - Thomson Reuters

6 Data refining APIs metadata and converting Annual report cover image Microsoft Cognitive API Feature Value Tags { "name": "person", "confidence": 0.998}, { "name": "man", "confidence": 0.934}, { "name": "indoor", "confidence": 0.929} Faces Color Image size 2690 x 2339 Adult content Racy score (Ethnicity) { "age": 20, "gender": "Male } {"background": grey, "foreground": red} Indian Example APIs: - Face to demographics - Voice to gender - Name to ethnicity - Report to entities - Article to summary - Video to annotation - Speech to text - Image to description - Text to tags

7 Intelligence APIs teach machine to classify Training data Testing data Three Musketeers Moby Dick Crime and Punishment Count de Rochefort Ishmael Dmitri Razumikhin Monsieur Bonacieux Captain de Deer Andrei Lebezyatnikov Duke of Buckingham Dough Boy Nastasya Petrovna Bazin Starbuck Porfiry Petrovich Felton Ahab Katerina Marmeladov Anne of Austria Flask Pyotr Luzhin Athos Fedallah Pulcheria Raskolnikov Kitty Stubb Alexander Zamyotov Constance de Bonacieux Captain Bildad Ilya Petrovich Planchet Queequeg Rodion Raskolnikov Monsieur de Tréville Daggoo Lizaveta Ivanovna Milady de Winter Elijah Sofya Marmeladov Porthos Captain Peleg Alyona Ivanovna IBM Watson Character Which book? Accuracy D'Artagnan Three musketeers? True Grimaud Three musketeers? True Aramis Three musketeers? True Mousqueton Three musketeers? True Captain Boomer Moby? Dick True Father Mapple Moby? Dick True Tashtego Moby? Dick True Pip Moby? Dick True Semyon Marmeladov Crime and? punishment True Arkady Svidrigailov Crime and? punishment True Zossimov Moby? Dick False Avdotya Raskolnikov Crime and? punishment True

8 Analysis APIs insights from non-numerical data Emotions Microsoft Face API Personality IBM Watson Personality Insights Concepts IBM Watson Natural Language Understanding anger: 0.00, contempt: 0.00, disgust: 0.00, fear: 0.00, happiness: 0.73, neutral: 0.26, sadness: 0.00 surprise: 0.00 This is my rifle. There are many like it, but this one is mine. My rifle is my best friend. It is my life. I must master it as I must master my life. My rifle, without me, is useless... You are innerdirected, skeptical and strict. Introvert: 0.99 Dutifulness: 0.73 Authoritychallenging: 0.39 Agreeableness:0.02 Our Father, who art in heaven, hallowed be thy name, thy kingdom come, thy will be done, on earth as it is in heaven. Give us this day our daily bread and forgive us our debts Christianity: 0.95 Christian prayer: 0.79 Christian terms: 0.78 American novels: 0.66 Lord's Prayer: 0.60 Me! Rifleman s Creed Lord s Prayer Example APIs: face to emotion, speech to emotion, diary to personality, story to concepts, conversation to tone, speech to person

9 Scaling up literary research For-loop that sends textual material to API: For Each Book in Library: For Each Page in Book: Result(Page) <= Send(Page, API) Next Page Next Book Cumulative Standardized Sentiment Fritz Pfeffer joins the annex Day-by-day sentiment of Anne Frank ( ) Library none Book The Diary of a Young Girl API IBM Watson Sentiment Analysis The sun is shining I think spring is inside me. I feel spring awakening

10 Scaling up communication studies For-loop that sends images to API: For Each Movie in Collection: For Each Frame in Movie: Result(Frame) <= Send(Frame, API) Next Frame Next Movie Expressed emotions in Trump-Clinton Debate Collection none Movie Final presidential debate API Microsoft Emotion API Trump s primary facial expression is angry, and more so if not having the floor Hillary emphasizes message by raising eyebrows, expresses integrity by smiling

11 Key takeaways - Non-numeric data can be obtained and analyzed at an industrial scale - Research workflow can be automated by chaining APIs cleverly - Huge implications on what can be analyzed and at what cost