Demographic-aware word associations

Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing, pages 2285–2295 Copenhagen, Denmark, September 7–11, 2017. c⃝2017 Association for Computational Linguistics Demographic-Aware Word Associations Aparna Garimella, Carmen Banea, and Rada Mihalcea University of Michigan Ann Arbor, MI {gaparna,carmennb,mihalcea}@umich.edu Abstract Variations of word associations across dif- ferent groups of people can provide in- sights into people’s psychologies and their world views. To capture these variations, we introduce the task of demographic- aware word associations. We build a new gold standard dataset consisting of word association responses for approximately 300 stimulus words, collected from more than 800 respondents of different gender (male/female) and from different locations (India/United States), and show that there are significant variations in the word as- sociations made by these groups. We also introduce a new demographic-aware word association model based on a neu- ral net skip-gram architecture, and show how computational methods for measur- ing word associations that specifically ac- count for writer demographics can outper- form generic methods that are agnostic to such information. 1 Introduction Understanding the associations that are formed in the mind is paramount to understanding the way humans acquire language throughout a lifetime of learning (Elman et al., 1997; Rogers and McClel- land, 2004). Furthermore, word associations are believed to mirror the mental model of the concep- tual connections in a human mind, and constitute a direct path to assessing one’s semantic knowledge (Nelson et al., 2004; Mollin, 2009). Word associations start forming early in life, as language is acquired and one learns based on the environment where concepts lie in relation to each other. For example, we may learn to associate “mother” with “warmth,” or “fire” with “burn.” Yet, this mental model is not static but highly dy- namic, and is shaped by new experiences over a lifetime. For instance, (Tresselt and Mayzner, 1964) showed that word associations change with time, and that for respondents in younger age groups their variability is lower, while for those in older age groups the variability is higher, as their life experiences modify the commonality between respondents from the same group. Computational linguistics has traditionally taken the “one-size-fits-all” approach, with most models being agnostic to the language of the speakers behind the language. With the introduc- tion and adoption of Web 2.0, there has been an exponential increase in the availability of digital user-centric data in the form of blogs, microblogs and other forms of online participation. Such data often times can be augmented with demographic or other user-focused attributes, whether these are user-provided (e.g., from a user’s online profile) or labeled using an automatic system. This enables computational linguists to go beyond generic corpus-based metrics of word associations, and attempt to extract associations that pertain to given demographic groups that would not have been possible without administering time consuming and resource intensive word association surveys. While current NLP methods generally deal with more advanced tasks (relation extraction, text sim- ilarity, etc.), at their very core many of these tasks assume some way …

Content truncated to 500 words