REAL TIME CROSS CORRELATION OF INTENSITY AND SENTIMENT FROM SOCIAL MEDIA MESSAGES

1 . A method for finding patterns in a target real-valued time series by utilizing sentiment and frequency derived from a stream of social media messages, wherein the target represents a quantifiable property of an asset being tracked, comprising:

identifying a target, which is a sampled real-valued time series;

generating a sentiment time series, s s , relating to an asset;

generating a frequency time series, s f , relating to an asset;

determining a pattern based upon the sentiment time series and the frequency time series.

2 . The method according to claim 1 , wherein sentiment is an expression of a psychological state relative to an event.

3 . The method according to claim 1 , wherein frequency represents the volume of social media messages about the asset.

4 . The method according to claim 1 , wherein the step of generating a sentiment time series is performed by language processing and is derived based upon pairs of lexical items in local syntactic context found in a volume of social media messages.

5 . The method according to claim 4 , wherein the step of generating a sentiment time series includes the creation of an average sentiment series, s a , such that for every point (t,s) in the sentiment time series, s s , there is a point (t, a) in an average sentiment series where “a” is the arithmetic average of all the sentiments in a time range [t−w, t].

6 . The method according to claim 5 , wherein the step of generating a sentiment time series includes the creation of a sentiment-frequency series, s sf , to contain a point (t,v sf ) for every (t, a) in the sentiment time series, s s , and (t, f) in the frequency time series, s f , where v sf =f a (=e a ln(f) ).

7 . The method according to claim 1 , wherein the frequency time series, s f , is dependent upon the sentiment time series, s s , and a positive number w representing a time called window size.

8 . The method according to claim 7 , wherein for each point (t, s) in the sentiment time series, s s , the frequency time series, s f , contains a point (t, f) where f is the number of points in the sentiment time series, s s , in the time range [t−w, t], divided by w.

9 . The method according to claim 8 , wherein the number f is called frequency and

f



(

t

)

=

#





(

s

s

⋂

[

t

-

w

,

t

]

)

w

s

f

=

{

(

t

,

f



(

t

)

)





t

∈

V

1



(

s

s

)

}

10 . The method according to claim 9 , wherein the pattern P is a cross-correlation c in [−1,1], a positive window size w, a time lag l, and a time t s , and these numbers are interpreted as a predictive series over [t s −w, t s ] correlating to the target series over [t s −w+l, t s +l] with a cross-correlation of c″.

11 . The method according to claim 1 , wherein the step of determining a pattern employs a sentiment-frequency method that uses sentiment to create a sentiment-frequency series, s fs , and correlates to the target using a plain statistical cross-correlation.

12 . The method according to claim 11 , wherein the step of determining a pattern includes the step of identifying an optimal time lag.

13 . The method according to claim 12 , wherein correlating two time-series using a plain statistical cross-correlation and finding the optimal lag is achieved with a series correlator.

14 . The method according to claim 13 , wherein the series correlator produces a set of patterns based on a real-valued pulsated time series s p , a real-valued sampled time series, s s , an interpolation method I for s s , and a window size w.

15 . The method according to claim 11 , wherein the interpolation method I, is a function of a time series s s and of a time t that is C 1 -piecewise continuous with respect to t, and such that if there exists a point (t, v) in s s , I(s s , t)=v.