# Understanding How People Use Natural Language to Ask for Recommendations: Query Dataset This dataset describes subjects' initial and follow-up queries from the research paper "Understanding How People Use Natural Language to Ask for Recommendations", published in the ACM Conference on Recommender Systems (RecSys), 2017. The data were collected on [movielens.org](http://movielens.org) between May 12 and May 24, 2016. This material is based on work supported by the National Science Foundation under grants IIS-0964695, IIS-1017697, IIS-1111201, IIS- 1210863, and IIS-1218826, and by a grant from Google. This readme was written by Max Harper on June 22, 2017. # Citation Jie Kang, Kyle Condi , Shuo Chang, Joseph A. Konstan, Loren Terveen, and F. Maxwell Harper. 2017. Understanding How People Use Natural Language to Ask for Recommendations. In Proceedings of RecSys ’17, Como, Italy, August 27-31, 2017, 9 pages. ## Contact Information If you have questions, contact Max Harper or ## License This work is licensed under a [Creative Commons Attribution 3.0 United States License](https://creativecommons.org/licenses/by/3.0/us/). # Description of responses.csv The dataset contains 347 rows, each row containing one subject's input. 224 of the subjects contributed survey responses, and 151 of the subjects contributed follow-up queries. ## Fields in responses.csv *firstQuery* -- The subject's first query, prompted by "I can help you find movies. What are you looking for?" *followUpQuery* -- The subject's follow-up query, prompted by "I can improve these results. Tell me more about what you want." *modality* -- The subject's modality for their first and follow-up queries. * `typing` -- the subject typed their query * `speaking` -- the subject spoke their query *assignmentCategory* -- * `typing-no-mic` -- the subject reported no microphone * `typing-random` -- the subject reported that they have a microphone and were randomly assigned to the `typing` modality * `typing-mic-not-working` -- the subject reported that they have a microphone, were randomly assigned to the `speaking` modality, then clicked "my microphone is not working" and were switched to the `typing` modality * `speaking-random` -- the subject reported that they have a microphone and were randomly assigned to the `speaking` modality *firstQueryTime* -- Number of seconds between opening the page and submitting the first query. *howOftenVoiceAssistant* -- Response to the survey question "How often do you use voice assistants (e.g., Google Now, Siri, etc.)?" * 0 -- Never * 1 -- Rarely * 2 -- A few times a month * 3 -- A few times a week * 4 -- Once a day * 5 -- Multiple times a day *otherVoiceAssistant* -- Response to the survey question "What other voice assistant or voice search tools have you used?" * Siri * Cortana * Alexa * Google Now * Others [Please specify] Empty brackets indicate that the subject chose none of the answers (i.e., they had not used any of the voice search services). Blank cells indicate that the subject did not provide survey responses. *shouldIncludeRated* -- Response to the survey question "For the queries you gave us, should we include movies you've already rated in the results or not?" * yes * no * unsure Blank cells indicate that the subject did not provide survey responses.