----------------------- REVIEW 1 ---------------------
PAPER: 540
TITLE: Understanding Human Movement Semantics: A Point of Interest Based Approach
AUTHORS: Ionut Trestian, Kévin Huguenin, Ling Su and Aleksandar Kuzmanovic

OVERALL RATING: 2 (accept)
REVIEWER'S CONFIDENCE: 3 (high)

The paper connects human trajectory data with certain points of interests (such as McDonalds restaurants or hospitals), by analyzing two existing sets of data: one from USA and one from China. The view of that of analyzing the movement patterns of the whole population given in the data.

The paper is well written in flawless English.

The paper is interesting, but includes some surprising incorrect statements. Already in the abstract you say “all of the studies isolate movements from the environment that surrounds people, i.e., the points of interest that they visit”. This simply is not true. Work has been undertaken, see for instance the references given below, to infer points of interest from trajectory data. Also reference [36] in the present paper is such a paper linking movements to points of interest.

Later on in the paper, it is stated that this is the first-of-its-kind joint analysis of mobility and POIs. It may be so on the population scale, but as already noted, not on the scale of an individual user.

You could have said in Table 1 that the mass center figures refer to latitude and longitude.

In Section 2.2 you mention three ways of obtaining points of interest. A further fourth method would be automatic inference of POIs from trajectory data (see the same references already mentioned).

References to other relevant work (not currently included in the list of reference):

- Kim, D. H.; Kim, Y.; Estrin, D. & Srivastava, M. B.: SensLoc: sensing everyday places and paths using less energy, Proceedings of the 8th ACM Conference on Embedded Networked Sensor Systems (SenSys), ACM, 2010, 43-56

- Hightower, J.; Consolvo, S.; LaMarca, A.; Smith, I. & Hughes, J.: Learning and Recognizing the Places We Go, Proceedings of the 7th International Conference on Ubiquitous Computing (UBICOMP), Springer-Verlag, 2005, 3660, 159-176

- Ashbrook, D. & Starner, T.: Using GPS to learn significant locations and predict movement across multiple users, Personal and Ubiquitous Computing, 2003, 7, 275 - 286

- Marmasse, N. & Schmandt, C.: Location-Aware Information Delivery with ComMotion, Proceedings of the 2nd International Symposium on Handheld and Ubiquitous Computing (HUC), Springer, 2000, 1927, 361-370

- Nurmi, P. & Bhattacharya, S.: Identifying Meaningful Places: The Nonparametric Way, Proceedings of the 6th International Conference on Pervasive Computing (Pervasive), Springer, 2008, 5013, 111-127

- Zhou, C.; Frankowski, D.; Ludford, P.; Shekhar, S. & Terveen, L.: Discovering Personally Meaningful Places: An Interactive Clustering Approach, ACM Transactions on Information Systems, 2007, 25, 12


Overall evaluation: This is an interesting and well-written paper. The unfounded claims need to be edited before publication. See also the other comments.


----------------------- REVIEW 2 ---------------------
PAPER: 540
TITLE: Understanding Human Movement Semantics: A Point of Interest Based Approach
AUTHORS: Ionut Trestian, Kévin Huguenin, Ling Su and Aleksandar Kuzmanovic

OVERALL RATING: 0 (borderline paper)
REVIEWER'S CONFIDENCE: 2 (medium)

This paper describes an approach to analyze human movement to understand its semantics,
based on the analysis of GPS mobility traces. The study is done (and compared) through
two different datasets: one in USA (large number of users, but limited in time and space)
and one in China (medium number of users, but long period of time and over all the country).

The originality of the work is not very high. This subject has been studied a lot in the
past few years. For example in:
+ "Unveiling the complexity of human mobility by querying and mining massive trajectory data",
Giannotti et al. (VLDB Journal, 2011. http://dl.acm.org/citation.cfm?id=2039170)
+ "Mining interesting locations and travel sequences from GPS trajectories", Zheng et al., WWW'09. (in the paper's references: [36])
The specificity of the paper is to correlate point of interests as explanations and observed patterns. An interesting aspect is the fact of using two datasets of culturally different countries.

Hypothesis at the beginning of the paper seems to state the obvious. For ex. "we believe that most human movement is driven by certain purposes." Yes, even ants movements have driven by goals...
This is also obvious that people are going more often to restaurants than to stadiums.
A dimension not taken into account into the study is cost: cost of movement (not important - or mandatatory for daily routine, but more important in other cases), and cost related to a point of interest (not meaningful for a supermarket, but different in a restaurant or a stadium - this might be somehow correlated to frequency of visit of that POI).

The study of point of interests brings information regarding human movements, but what's about
the analysis of social ties, as adressed for example in "Human Mobility, Social Ties, and Link Prediction" (Wang et al., KDD'2011)?
http://www.barabasilab.com/pubs/CCNR-ALB_Publications/201108-21_KDD-HumanSocialTies/201108-21_KDD-HumanSocialTies.pdf
I think this is quite difficult to ignore it as it relates also to potential points of interest. In particular, the study is limited to available points of interests, i.e. those well known to everybody such as the Eiffel Tower or the Empire State Building. But what's about "personal" points of interests, such as the one that guides my social and personal behaviour, such as: the house of my oncle, the place where my best friends live, etc.
which are very important for me as individual?

My main concern about the paper is that although the work is relatively serious, the hypothesis and conclusions are quite obvious (e.g. the authors are surprised by the low frequency of visits to libraries), not surprising, and can certainly be validated by less complex means (for example just a questionnaire as done in sociology).
So, if the goal was to confirm with usage data known facts in social sciences (such as sociology and psychology), then, this is fine, but the paper should be presented accordingly.
I would have loved to see for ex. the discovery of a new point of interest, unknown from the map or something like that.

More interesting to me is for example the work of Qiu (http://www.cigi.illinois.edu/cybergis/docs/Qiu_Position_Paper.pdf)
which studies mobility movements in relationship with some "specific" POI: endemic places of malaria in Tanzania...

So, overall, the approach and the methodology is relatively interesting, but should tackle more specific problems (maybe by changing the time window used for the analysis?)

Some details:
- section 2.1: the two datasets are very different. How the American one (1 week only) can be used do demonstrate behaviours that are nor pure routine?
- p3: why restricting to a few types of POI?
- section 5.1: for user interests, is _time_ spent at a POI taken into account? it's not because I passed by McDonald everyday that I like it... In particular in big cities where the density of POIs is high, many of them are not relevant for the user.
The results of 5.1 prove well-known cultural differences between USA and China, which is good
- section 5.3: the energy study is interesting. But, instead of looking at the differences between categories, which are quite obvious, I would have been more interested to see the potential divergences within a category: ex. why a specific McDonald attracts more (higher energy) than the average in this category. (ex. maybe the colocation with another POI?)


----------------------- REVIEW 3 ---------------------
PAPER: 540
TITLE: Understanding Human Movement Semantics: A Point of Interest Based Approach
AUTHORS: Ionut Trestian, Kévin Huguenin, Ling Su and Aleksandar Kuzmanovic

OVERALL RATING: 2 (accept)
REVIEWER'S CONFIDENCE: 1 (low)

The paper presents an interesting idea and seems to be backed by a lot of good analysis.

It is not clear as to how valuable the conclusions presented in this paper are likely to be for businesses. For example, in order to do targeted advertising, business owners in many cases already have info regarding the time of day behavior and proximity of a user base to their location. This could however be of some value to upcoming businesses or to keep track of overall changes in trends. It would helpful if the authors could clarify the utility of their conclusions by providing specific examples on how they add value to businesses (keeping in mind that business owners already have some data at their disposal today).


----------------------- REVIEW 4 ---------------------
PAPER: 540
TITLE: Understanding Human Movement Semantics: A Point of Interest Based Approach
AUTHORS: Ionut Trestian, Kévin Huguenin, Ling Su and Aleksandar Kuzmanovic

OVERALL RATING: 0 (borderline paper)
REVIEWER'S CONFIDENCE: 3 (high)

Meta review for submission #540:

Contribution: the paper presents an approach to analyze movement of mobile uses to understand its semantics. The presented method is based on the analysis of GPS traces. The study is done (and compared) through
two different datasets, one in USA (large number of users, but limited in time and space) and one in China (medium number of users, but long period of time and over all the country).

Quality/fit: reasonable paper in quality and presentation. Good fit to call for papers as contribution in the area of location-based services and context awareness.

Weakness: the main concern about the paper is that although the work is relatively serious, the hypothesis and conclusions are rather obvious, not surprising. Subsequently, extensive related literature is pointed out by Reviews #1--3.