Archive for related work

The Language of Location

 

During my work, I've often noticed similarities between language and individual daily life location behaviour (as detected by GPS, cell towers, tweets, check-ins etc.). To summarise these thoughts, I've compiled a list of the similarities and differences between language and location below. I then mention a few papers that exploit these similarities to create more powerful or interesting approaches to analysing location data.

Similarities between Location and Language Data

  • Both exhibit power laws. A lot of words are used very rarely while a few words are very frequently used. The same happens with the frequency of visits to locations (e.g., how often you visit home v.s. your favourite theme park). This is not a truism. The most frequently visited locations or words used are *much* more likely to be visited/used than most other places/words.
  • Both exhibit sequential structure. Words are highly correlated with words near to them on the page. The same for locations on a particular day.
  • Both exhibit topics or themes. In the case of language, groups of words tend to co-occur in the same document (e.g., two webpages that talk about cars are both likely to mention words from a similar group of words representing the "car" topic). In the case of location data, a similar thing happens. I mention two interpretations from specific papers later in this post.
  • The availability of both language data and location data has exploded in the last decade (the former from the web, the latter from mobile devices).
  • There are cultural differences in using language just as there are cultural differences in location behaviour (e.g., Spanish people like to eat out later than people of other cultures).
  • Both are hierarchical. Languages have letters, words, sentences, and paragraphs. A person can be moving around at the level of the street, city, or country (during an hour, day, or week).
  • Both exhibit social interactions. Language is exchanged in emails, texts, verbally, or in scholarly debate. Friends, co-workers, and family may have interesting patterns of co-location.

Differences between Location and Language Data

  • Many words are shared between texts (of same language) but locations are usually highly personal to individuals (except for the special cases of friends, co-workers, and family).
  • There are no periodicities in text but strong periodicities in location (i.e., hourly, weekly, and monthly).
  • Language data is not noisy (except for spelling and grammar mistakes) while location data is usually noisy.
  • Language analysts do not usually need to worry about privacy issues whilst location analysts usually do.

Work that Exploits These Similarities

Here are a few papers that apply or adapt approaches that were primarily used for language models to location data:

K. Farrahi and D. Gatica-Perez. Extracting mobile behavioral patterns with the distant n-gram topic model. In Proc. ISWC, 2012.

They use topic modelling to capture the tendency of visiting certain locations on the same day. This is similar to using the presence of words like "windshield" and "wheel" to place higher predictive density on words like "road" and "bumper" (i.e., topic modelling bags of words). I have talked previously about why I think this is a good paper.

L. Ferrari and M. Mamei. Discovering daily routines from google latitude with topic models. In PerCom Workshops, pages 432–437, 2011.

This paper uses a similar application of topic modelling as the one by Farrahi and Gatica-Perez.

H. Gao, J. Tang, and H. Liu. Exploring social-historical ties on location-based social networks. In 6th ICWSM, 2012.

This paper uses a model that was previously used to capture sequential structure in words and applies it to Foursquare checkins.

J. McInerney, J. Zheng, A. Rogers, N. R. Jennings. Modelling Heterogeneous Location Habits in Human Populations for Location Prediction Under Data Sparsity. In Proc. UbiComp, 2013.

In my own work, I've used the concept of topics to refer to location habits that represent the tendency of an individual to be at a given location at a certain time of day or week. This way of thinking about locations is useful in generalising temporal structure in location behaviour across people, while still allowing for topics/habits to be present to greater or varying degrees in different people's location histories (just as topics are more or less prevalent in different documents).

Both language and location data are results of human behaviour, so it is unsurprising to find similarities, even if I think some of the similarities are coincidental (e.g., power laws crop up in many places and often for different reasons, and the increasing availability of data is part of the general trend of moving the things we care about into the digital domain). The benefits of analysis approaches seem to be flowing in the language -> location direction only at the moment, though I hope one day that will change.

Top 5 Papers for Mobile Location Prediction

 

Below are some of the most useful papers for location prediction from mobile data (GPS, cell towers, Foursquare checkins):

  • Ashbrook, Daniel, and Thad Starner. "Using GPS to learn significant locations and predict movement across multiple users." Personal and Ubiquitous Computing 7.5 (2003): 275-286.
    • A classic in the field, introducing a vision for finding significant places (with respect to individuals) and learning structure from past behaviour to make predictions. Although there are some shortcomings (e.g., their method of finding significant locations involved manual calibration, and they only considered a first-order Markov model to make predictions), I think they provide a compelling vision that stands the test of time 10 years later.
  • Eagle, Nathan, and Alex Sandy Pentland. "Eigenbehaviors: Identifying structure in routine." Behavioral Ecology and Sociobiology 63.7 (2009): 1057-1066.
    • In this paper, Eagle and Pentland represent each day of an individual's mobility as a point in high dimensional space, then use dimensionality reduction (principle component analysis) to find a set of "eigenbehaviors" that best characterises the location behaviour of that individual. Prediction can then be done by finding the mix of eigenbehaviors that best recreates a partially-seen day. The general form of the eigenbehaviors also allows comparison of habits between people, and they found some nice results showing how students from different faculties have similar location habits. I prefer the exploratory applications of this approach more than the predictive aspect, because I think the high accuracy results they report are more a function of people staying a long time at the same location in their dataset (since they consider a person to always be in 1 of 4 locations: home, work, other, or no signal). Still, I was always inspired by this paper. An extension that considered richer data sets was done by Sadilek and Krumm (2012), which found good results when also incorporating temporal information (day of week, whether the day was a national holiday).
  • Gao, Huiji, Jiliang Tang, and Huan Liu. "Exploring Social-Historical Ties on Location-Based Social Networks." ICWSM. 2012.
    • I think this is a strong paper that deserves more attention. It is an extension of the Ashbrook and Starner paper in the sense that the authors provide a more sophisticated way of doing sequential location prediction (i.e., "given your recent locations, where are you likely to be next?") using hierarchical Pitman-Yor processes (HPY). HPY assumes that the pattern of individual daily life locations follows a power law, with the rich getting richer (i.e., there are a few locations that are highly visited, and lots of locations that are hardly ever visited), which has been empirically observed in Gonzalez et al. (2008). Furthermore, HPY is a Bayesian way of varying the number of recent locations considered in the prediction (similar to variable-order Markov models, but with a more principled motivation that doesn't require a max-depth parameter). I suggest reading the paper for a more detailed description than I can give here.
  • Horvitz, Eric, and John Krumm. "Some help on the way: Opportunistic routing under uncertainty." Proceedings of the 2012 ACM Conference on Ubiquitous Computing. ACM, 2012.
    • The authors propose an anticipatory mobile computing system that recommends interesting places to drivers on their way to a predicted location. The prediction part of the system is assumed (using a previous approach that assumes a rational driver) so the focus is on how to calculate the places to recommend using expected utility. They also consider how to balance the expected benefit of asking a driver to confirm their destination (making the suggested places more relevant) against the cost of interruption. Clearly, there will be times when knowing the user's destination for certain would not change the recommendation very much (v.s. only having a probabilistic prediction), so this approach is useful in avoiding the "talking paperclip" syndrome where anticipatory applications interrupt users too much.
  • Farrahi, Katayoun, and Daniel Gatica-Perez. "A probabilistic approach to mining mobile phone data sequences." Personal and Ubiquitous Computing (2013): 1-16.
    • This work provides a Bayesian probabilistic model of individual location behaviour that is a hybrid of latent Dirichlet allocation (LDA) and the eigenbehaviors approach of Eagle and Pentland. The similarity to eigenbehaviors comes from the assumption that there exists a set of characteristic motifs that repeat themselves in the data (e.g., leaving home to go to the bus station, then to work). The authors' comparison to n-gram models appears to be a bit of a red herring to me, as the next location in their model is not dependent, in general, on the most recent previous locations (that is not to say a mobility model requires the n-gram assumption to be useful). The benefit of having LDA underneath it all (as opposed to, say, a mixture model) is to express the assumption that motifs are locally correlated within a single day. Intuitively, if I follow the aforementioned "going to work" motif in the morning, then I am probably more likely to follow other workday-related motifs (e.g., going to the gym, then home) than other types of motif later that day. With hierarchical Bayesian modelling, this type of structure can be learnt in an unsupervised way from the data, and then be used to make predictions about future behaviour.

I had to leave a lot of very good papers out, so I hope to make a longer list in future.