220 likes | 320 Views
Advisor: Hsin-Hsi Chen Speaker: Sheng-Chung Yen Date: 2007/04/09. Of Men, Women, and Computer: Data Driven Gender Modeling for Improved User Interfaces. Hugo Liu, Rada Mihalcea MIT, University of North Texas. Introduction.
E N D
NTU Natural Language Processing Lab. Advisor: Hsin-Hsi Chen Speaker: Sheng-Chung Yen Date: 2007/04/09 Of Men, Women, and Computer:Data Driven Gender Modeling for Improved User Interfaces Hugo Liu, RadaMihalcea MIT, University of North Texas
NTU Natural Language Processing Lab. Introduction • Men and Women think and feel differently, and perceive, value and understand the world in their own ways. • Contributions: • They describe a corpus-based approach to gender modeling. • They build GENGERLENS - a novel intelligent news filtering system that customizes news based on the gender of its reader.
NTU Natural Language Processing Lab. Data • Corpus • Blogspot, LiveJournal, and MSN-Space • 75000 male blog entries and 75000 female blog entries • With blogger profile • 2006.07.27 or 28
NTU Natural Language Processing Lab. Feature Scoring
NTU Natural Language Processing Lab. Dimensions of the gender space • Time • Food • Color • Size • Socialness • Affect
NTU Natural Language Processing Lab. Time • relative-time expressions • such as ”last week” • concrete-time expressions • such as ”Wednesday”
NTU Natural Language Processing Lab. • Women : here-and-now, from ”last weekend” through to ”this weekend.” • Men are more likely to focus on events of the past and future months and years. • Feminine writing dominates the days-of-the-week. • Masculine writing prefers to focus on months-of the-year.
NTU Natural Language Processing Lab. Food • Their experiment was to utilize the ontology of food terms from WordNet. • Feminine: Sweets and healthy foods • Masculine: liquids and hearty foods • Women paid more attention to the details of food.
NTU Natural Language Processing Lab. Color • They started with the widely-used X11 color lexicon.(http://en.wikipedia.org/wiki/List_of_colors) • color order – a concept in color theory which prescribes every color as being either primary, secondary, tertiary (3rd order), quaternary (4th order), and so on.
NTU Natural Language Processing Lab. Size • They generated five size graded expressions for each word. • For example, the feminine feature “skirt” generated the terms: “tiny skirt,” “small skirt,” “average skirt,” “big skirt,” “huge skirt.”
NTU Natural Language Processing Lab. Socialness • The results of the experiment found • relative3 (aunty, sibling, and groom) saw an average orientation of 0.16, thus leaning toward the feminine; • socialgroup1 (staff, church, and bikers) saw an average orientation of -0.22, thus leaning toward the masculine.
NTU Natural Language Processing Lab. Affect • ANEW – a set of normative affective ratings for 1034 common English words. • ANEW rates words using the pleasure-arousal-dominance (PAD) model of emotion. • PAD model • (P)leasure ranges from joy (+P) to reluctance (-P) • (A)rousal ranges from mental awareness (+A) to sleepiness (-A) • (D)ominance describes the agent‘s feelings of control over the situation
NTU Natural Language Processing Lab. • Pleasuremale = 0.047; Pleasurefemale = 0.096 • Arousalmale = 0.048; Arousalfemale = 0.014 • They opted for Ekman’s ontology of six universal emotions + ’neutral. • Surprise, disgusted, fearful, angry, sad, happy + neutral
NTU Natural Language Processing Lab. Gender Lens • GENDERLENS • reading the news feed from a major news aggregator (Google News). • a news filtering system that reranks the daily news based on the gender biases learned from the blog data set.
NTU Natural Language Processing Lab. Evaluation • 30 users (15 men and 15 women)