Dynamic modeling of geospatial words in social media

Patent metadata

Field	Value
Publication number	US-9405743-B1
Application number	US-201514710915-A
Country	US
Kind code	B1
Filing date	May 13, 2015
Priority date	May 13, 2015
Publication date	Aug 2, 2016
Grant date	Aug 2, 2016

Abstract

Dynamically modelling geospatial words in social media, in one aspect, generates a word set based on frequencies of words occurring in GPS annotated text data generated by a GPS-enabled device containing latitude and longitude coordinates. Locations are partitioned by mapping GPS coordinates in the GPS annotated text data to a set of discrete non-overlapped locations. A text stream contained in the GPS annotated text data is segmented into time windows. Footprints of locations in time windows are generated. Geospatial weights associated with words in the word set are generated based on localness of words determined based on the footprints. Words in a text message are extracted and scores are determined for the set of discrete non-overlapped locations associated with the words.

First claim

We claim:

1. A system for dynamically modeling geospatial words in social media, comprising:
a processor;
a data collector operable to execute on the processor and further operable to receive GPS annotated text data generated by a GPS-enabled device containing latitude and longitude coordinates;
a model trainer operable to execute on the processor and further operable to generate a word set based on frequencies of words occurring in the GPS annotated text data,
the model trainer further operable to partition locations by mapping GPS coordinates in the GPS annotated text data to a set of discrete non-overlapped locations,
the model trainer further operable to segment a text stream contained in the GPS annotated text data into time windows,
the model trainer further operable to generate footprints of locations in time windows,
the model trainer further operable to determine geospatial weights associated with words in the word set based on localness of words determined based on the footprints,
the model trainer further operable to dynamically integrate geotagging by extracting words in a text message and determining scores associated with the set of discrete non-overlapped locations;
a storage device coupled to the processor and operable to store the footprints and GPS labeled data, the GPS labeled data generated based on mapping the words in the word set to a respective location in the set of discrete non-overlapped locations.

2. The system of claim 1 , further comprising a model deployer operable to execute on the processor and further operable to predict location information for a new text message based on words in the new text message and the geospatial weights.

3. The system of claim 1 , wherein the model trainer samples a fixed number of the time windows in segmenting the text stream into time windows and generating the footprints.

4. The system of claim 1 , wherein the model trainer generates footprints of locations in time windows by, for each GPS annotated text data in a time window, constructing a bipartite graph between a word type of the GPS annotated text data and a mapped location.

5. The system of claim 4 , wherein the model trainer generates footprints by further determining an association strength between the word type and the mapped location.

6. The system of claim 5 , the model trainer further selects a number of locations based on the association strength as the footprints, the footprints parameterized by associated word type, time window, and the number.

7. A non-transitory computer readable storage medium storing a program of instructions executable by a machine to perform a method of dynamically modeling geospatial words in social media, the method comprising:
receiving GPS annotated text data generated by a GPS-enabled device containing latitude and longitude coordinates;
generating a word set based on frequencies of words occurring in the GPS annotated text data;
partitioning locations by mapping GPS coordinates in the GPS annotated text data to a set of discrete non-overlapped locations;
segmenting a text stream contained in the GPS annotated text data into time windows;
generating footprints of locations in time windows;
determining geospatial weights associated with words in the word set based on localness of words determined based on the footprints;
dynamically integrating in geotagging by extracting words in a text message and determining scores associated with the set of discrete non-overlapped locations.

8. The non-transitory computer readable storage medium of claim 7 , further comprising:
sampling a fixed number of the time windows in the segmenting and the generating steps.

9. The non-transitory computer readable storage medium of claim 7 , wherein the generating footprints of locations in time windows comprises, for each GPS annotated text data in a time window, constructing a bipartite graph between a word type of the GPS annotated text data and a mapped location.

10. The non-transitory computer readable storage medium of claim 9 , wherein the generating footprints of locations further comprises determining an association strength between the word type and the mapped location.

11. The non-transitory computer readable storage medium of claim 10 , further comprising selecting a number of locations based on the association strength as the footprints, the footprints parameterized by associated word type, time window, and the number.

12. The non-transitory computer readable storage medium of claim 7 , further comprising predicting location information for a new text message based on words in the new text message and the geospatial weights.

Assignees

• IBM

Inventors

• Butler Christopher J
• Han Bo
• Lai Jennifer C

Classifications

• G06F16/9537 Primary

  Spatial or temporal dependent retrieval, e.g. spatiotemporal queries · CPC title

• G06F40/169

  Annotation, e.g. comment data or footnotes · CPC title

• G06F40/263

  Language identification · CPC title

• G06F16/35

  Clustering; Classification · CPC title

• G06F40/211

  Syntactic parsing, e.g. based on context-free grammar [CFG] or unification grammars · CPC title