Methods and apparatus for sentiment analysis
Analysis is enabled, of a corpus of statements (such as those from social media), according to each statement's expression of sentiment about some kind of object. Object-specific corpuses are identified, where each object-specific corpus contains statements that refer to a same object. For each statement of an object-specific corpus, the polarity and intensity of sentiment expressed is determined. Net polarity and intensity measures are determined for each object-specific corpus and utilized to graph the corpus in a two-dimensional space. The area of the graphical symbol, representative of an object-specific corpus, can be proportional to the number (absolute or relative) of statements of the object-specific corpus. Brands can be compared, with each brand represented by an object-specific corpus. A single brand can have shown, relative to a temporal dimension, the net polarity, net intensity, or volume of its statements. Net polarity is shown to have a strong correlation with survey-based techniques.
1. A method for graphically characterizing statements about an object, comprising:
applying, as a result of computing hardware and programmable memory, frame extraction, to a first corpus, in order to attempt to identify, for each statement of the corpus, an object and a sentiment expressed about the object;
identifying, as a result of computing hardware and programmable memory, a first object-specific corpus, that is a subset of the first corpus, where all the statements of the first object-specific corpus are about a same first object;
categorizing, as a result of computing hardware and programmable memory, a sentiment of each statement, of the first object-specific corpus;
categorizing a polarity of each statement, of the first object-specific corpus;
categorizing an intensity of each statement, of the first object-specific corpus;
determining, as a result of computing hardware and programmable memory, a net polarity measure as a function of the polarity categorization and a net intensity measure as a function of the intensity categorization;
producing, as a result of computing hardware and programmable memory, a first graphical representation, of the first object-specific corpus;
placing the first graphical representation, relative to a first axis, in accordance with the net polarity measure; and
placing the first graphical representation, relative to a second axis, in accordance with the net intensity measure.
2. The method of claim 1 , wherein the step of producing a first graphical representation further comprises:
producing a graphical representation having an area proportional to a number of statements contained in the first object-specific corpus.
3. The method of claim 1 , wherein the categorizing further comprises:
categorizing the polarity as either positive or negative; and
categorizing the intensity as either strong or weak.
4. The method of claim 1 , further comprising:
production of a first and a second dividing line, each dividing line parallel to, respectively, the first axis and the second axis.
5. The method of claim 4 , wherein production, of the first and second dividing lines, further comprises:
placing the first and second dividing lines in order to divide the area, for placement of the first graphical representation of the first object-specific corpus, into quadrants.
6. The method of claim 4 , further comprising:
applying frame extraction, to a second corpus, in order to attempt to identify, for each statement of the corpus, an object and a sentiment expressed about the object;
identifying a first plurality of object-specific corpuses, each of which is a subset of the second corpus, where all the statements, of an object-specific corpus, are about a same object;
categorizing a polarity and intensity of each statement of each object-specific corpus, wherein each object-specific corpus is a member of the first plurality of object-specific corpuses;
determining a second net polarity measure and a second net intensity measure, for each object-specific corpus, as a function of the categorization, wherein each object-specific corpus is a member of the first plurality of object-specific corpuses;
determining a median net polarity measure, from the second net polarity measures;
determining a median net intensity measure, from the second net intensity measures; and
placing each of the first and second dividing lines, respectively, at a location that approximately intersects the median net intensity and median net polarity measures.
7. The method of claim 6 , wherein the second corpus is the same as the first corpus and the first object-specific corpus is a member of the first plurality of object-specific corpuses.
8. The method of claim 4 , further comprising:
identifying a first plurality of object-specific corpuses, each of which is a subset of the first corpus, where all the statements, of an object-specific corpus, are about a same object, wherein each object-specific corpus is a member of the first plurality of object-specific corpuses;
categorizing the polarity and intensity of each statement of each object-specific corpus, wherein each object-specific corpus is a member of the first plurality of object-specific corpuses;
determining a net polarity measure and a net intensity measure, for each object-specific corpus, as a function of the categorization, wherein each object-specific corpus is a member of the first plurality of object-specific corpuses;
selecting a first subset of the first plurality of object-specific corpuses, wherein the first subset can contain all members of the first plurality of object-specific corpuses; and
placing the first and second dividing lines in order to divide an area, relative to the first subset, into quadrants for placement of a graphical representation for each member of the first subset.
9. The method of claim 8 , wherein each object, for each of the first plurality of object-specific corpuses, represents a different brand but all the brands are part of a same category of brand.
10. The method of claim 1 , further comprising:
applying frame extraction, to a second corpus, in order to attempt to identify, for each statement of the corpus, an object and a sentiment expressed about the object;
identifying a first plurality of object-specific corpuses, each of which is a subset of the second corpus, where all the statements, of an object-specific corpus, are about a same object;
categorizing a sentiment of each statement of each object-specific corpus of the first plurality of object-specific corpuses;
determining a net metric, for each object-specific corpus of the first plurality of object-specific corpuses, as a function of the categorization; and
determining a first median net metric measure, from the net metrics of the first plurality of object-specific corpuses; and
placing a first median dividing line, perpendicular to an axis, and located at, approximately, the first median net metric measure.
11. The method of claim 1 , wherein the first object represents a brand.
12. The method of claim 1 , further comprising:
producing a second corpus, containing statements within a second temporal range that is different from a first temporal range of the statements of the first corpus;
applying frame extraction, to the second corpus, in order to attempt to identify, for each statement of the corpus, an object and a sentiment expressed about the object;
identifying a second object-specific corpus, that is a subset of the second corpus, where all the statements of the second object-specific corpus are about the same first object as the first object-specific corpus;
categorizing the sentiment of each statement, of the second object-specific corpus;
determining a net metric value as a function of the categorization of the second object-specific corpus; and
producing a second graphical representation, of the second object-specific corpus, that is placed according to the second net metric.
13. The method of claim 12 , wherein the second temporal range is disjoint from the first temporal range.
14. The method of claim 12 , wherein a beginning and ending time, of the first and second temporal ranges, are the same.
15. The method of claim 12 , wherein producing a second graphical representation further comprises:
updating a position of the first graphical representation.
16. The method of claim 12 , wherein producing a second graphical representation further comprises:
producing a graphical representation additional to the first graphical representation.
17. The method of claim 12 , wherein producing a second graphical representation further comprises:
placing the second graphical representation relative to a temporal axis, with the first graphical representation occupying a differing position relative to the temporal axis.
18. The method of claim 1 , further comprising:
determining the net polarity measure from a first object-specific corpus that contains statements mostly from social media.
19. The method of claim 1 , wherein the net polarity measure has a strong correlation with a survey-based technique for determining sentiment.
20. The method of claim 19 , wherein the strong correlation corresponds to a value of 0.7 or greater, for a correlation coefficient.
21. The method of claim 1 , wherein the net polarity and intensity measures have a very low correlation.
22. The method of claim 21 , wherein the very low correlation corresponds to a value of 0.1 or less, for a correlation coefficient.
23. The method of claim 1 , wherein the net polarity measure has a very low correlation to a number of statements contained in the first object-specific corpus.
24. The method of claim 23 , wherein the very low correlation corresponds to a value of 0.1 or less, for a correlation coefficient.
25. The method of claim 1 , further comprising:
identifying a first plurality of object-specific corpuses, each of which is a subset of the first corpus, where all the statements, of an object-specific corpus, are about a same object and the first object-specific corpus is part of the first plurality of object-specific corpuses;
determining a total number of statements across the first plurality of object-specific corpuses;
producing a first relative value, of a first number of statements contained in the first object-specific corpus considered relative to the total number of statements; and
producing the first graphical representation of the first object-specific corpus with an area proportional to the first relative value.
26. The method of claim 25 , wherein the first relative value has a very low correlation to the net polarity measure.
27. The method of claim 26 , wherein the very low correlation corresponds to a value of 0.1 or less, for a correlation coefficient.
28. The method of claim 1 , further comprising:
identifying a first plurality of object-specific corpuses, including the first object-specific corpus, where each object-specific corpus is a subset of the first corpus;
limiting each corpus, of the first plurality of object-specific corpuses, such that all statements are about a same object and a same characteristic of the same object;
categorizing the polarity and intensity of each statement of each object-specific corpus;
determining a net polarity measure and a net intensity measure, for each object-specific corpus, as a function of the categorization;
placing a graphical representation, of each corpus of the first plurality of object-specific corpuses, relative to first and second axes, in accordance with, respectively, the net polarity and intensity measures;
dividing a graphing space, defined by the axes, into quadrants;
mapping each quadrant to representing one of strengths, weaknesses, opportunities and threats.
29. A system for graphically characterizing statements about an object, comprising:
a sub system configured, as a result of the computing hardware and programmable memory, to apply frame extraction, to a first corpus, in order to attempt to identify, for each statement of the corpus, an object and a sentiment expressed about the object;
a sub system configured, as a result of the computing hardware and programmable memory, to identify a first object-specific corpus, that is a subset of the first corpus, where all the statements of the first object-specific corpus are about a same first object;
a sub system configured, as a result of the computing hardware and programmable memory, to categorize a sentiment of each statement, of the first object-specific corpus;
a sub system configured, as a result of the computing hardware and programmable memory, to categorize a polarity of each statement, of the first object-specific corpus;
a sub system configured, as a result of the computing hardware and programmable memory, to categorize an intensity of each statement, of the first object-specific corpus;
a sub system configured, as a result of the computing hardware and programmable memory, to determine a net polarity measure as a function of the polarity categorization and a net intensity measure as a function of the intensity categorization;
a sub system configured, as a result of the computing hardware and programmable memory, to produce a first graphical representation, of the first object-specific corpus;
a sub system configured, as a result of the computing hardware and programmable memory, to place the first graphical representation, relative to a first axis, in accordance with the net polarity measure; and
a sub-system configured, as a result of the computing hardware and programmable memory, to place the first graphical representation, relative to a second axis, in accordance with the net intensity measure.
30. The system of claim 29 , further comprising:
a sub-system configured to apply frame extraction, to a second corpus, in order to attempt to identify, for each statement of the corpus, an object and a sentiment expressed about the object;
a sub-system configured to identify a first plurality of object-specific corpuses, each of which is a subset of the second corpus, where all the statements, of an object-specific corpus, are about a same object;
a sub-system configured to categorize a sentiment of each statement of each object-specific corpus of the first plurality of object-specific corpuses;
a sub-system configured to determine a net metric, for each object-specific corpus of the first plurality of object-specific corpuses, as a function of the categorization; and
a sub-system configured to determine a first median net metric measure, from the net metrics of the first plurality of object-specific corpuses; and
a sub-system configured to place a first median dividing line, perpendicular to an axis, and located at, approximately, the first median net metric measure.
31. The system of claim 29 , further comprising:
a sub-system configured to produce of a first and a second dividing line, each dividing line parallel to, respectively, the first axis and the second axis;
a sub-system configured to apply frame extraction, to a second corpus, in order to attempt to identify, for each statement of the corpus, an object and a sentiment expressed about the object;
a sub-system configured to identify a first plurality of object-specific corpuses, each of which is a subset of the second corpus, where all the statements, of an object-specific corpus, are about a same object;
a sub-system configured to categorize a polarity and intensity of each statement of each object-specific corpus, wherein each object-specific corpus is a member of the first plurality of object-specific corpuses;
a sub-system configured to determine a second net polarity measure and a second net intensity measure, for each object-specific corpus, as a function of the categorization, wherein each object-specific corpus is a member of the first plurality of object-specific corpuses;
a sub-system configured to determine a median net polarity measure, from the second net polarity measures;
a sub-system configured to determine a median net intensity measure, from the second net intensity measures; and
a sub-system configured to place each of the first and second dividing lines, respectively, at a location that approximately intersects the median net intensity and median net polarity measures.
32. The system of claim 29 , further comprising:
a sub-system configured to produce of a first and a second dividing line, each dividing line parallel to, respectively, the first axis and the second axis;
a sub-system configured to identify a first plurality of object-specific corpuses, each of which is a subset of the first corpus, where all the statements, of an object-specific corpus, are about a same object, wherein each object-specific corpus is a member of the first plurality of object-specific corpuses;
a sub-system configured to categorize the polarity and intensity of each statement of each object-specific corpus, wherein each object-specific corpus is a member of the first plurality of object-specific corpuses;
a sub-system configured to determine a net polarity measure and a net intensity measure, for each object-specific corpus, as a function of the categorization, wherein each object-specific corpus is a member of the first plurality of object-specific corpuses;
a sub-system configured to select a first subset of the first plurality of object-specific corpuses, wherein the first subset can contain all members of the first plurality of object-specific corpuses; and
a sub-system configured to place the first and second dividing lines in order to divide an area, relative to the first subset, into quadrants for placement of a graphical representation for each member of the first subset.
33. The system of claim 29 , further comprising:
a sub-system configured to produce a second corpus, containing statements within a second temporal range that is different from a first temporal range of the statements of the first corpus;
a sub-system configured to apply frame extraction, to the second corpus, in order to attempt to identify, for each statement of the corpus, an object and a sentiment expressed about the object;
a sub-system configured to identify a second object-specific corpus, that is a subset of the second corpus, where all the statements of the second object-specific corpus are about the same first object as the first object-specific corpus;
a sub-system configured to categorize the sentiment of each statement, of the second object-specific corpus;
a sub-system configured to determine a net metric value as a function of the categorization of the second object-specific corpus; and
a sub-system configured to produce a second graphical representation, of the second object-specific corpus, that is placed according to the second net metric.
34. The system of claim 29 , further comprising:
a sub-system configured to determine the net polarity measure from a first object-specific corpus that contains statements mostly from social media.
35. The system of claim 29 , wherein the net polarity measure has a strong correlation with a survey-based technique for determining sentiment.
36. The system of claim 29 , further comprising:
a sub-system configured to identify a first plurality of object-specific corpuses, each of which is a subset of the first corpus, where all the statements, of an object-specific corpus, are about a same object and the first object-specific corpus is part of the first plurality of object-specific corpuses;
a sub-system configured to determine a total number of statements across the first plurality of object-specific corpuses;
a sub-system configured to produce a first relative value, of a first number of statements contained in the first object-specific corpus considered relative to the total number of statements; and
a sub-system configured to produce the first graphical representation of the first object-specific corpus with an area proportional to the first relative value.
37. The system of claim 29 , further comprising:
a sub-system configured to identify a first plurality of object-specific corpuses, including the first object-specific corpus, where each object-specific corpus is a subset of the first corpus;
a sub-system configured to limit each corpus, of the first plurality of object-specific corpuses, such that all statements are about a same object and a same characteristic of the same object;
a sub-system configured to categorize the polarity and intensity of each statement of each object-specific corpus;
a sub-system configured to determine a net polarity measure and a net intensity measure, for each object-specific corpus, as a function of the categorization;
a sub-system configured to place a graphical representation, of each corpus of the first plurality of object-specific corpuses, relative to first and second axes, in accordance with, respectively, the net polarity and intensity measures;
a sub-system configured to divide a graphing space, defined by the axes, into quadrants; and
a sub-system configured to map each quadrant to representing one of strengths, weaknesses, opportunities and threats.