Methods and apparatus to classify text communications
View Patent ↗Methods and apparatus to classify text communications are disclosed. An example method includes determining a first conditional probability of a first feature occurring in a text given that the text belongs to a classification mode, wherein the first feature is included in the text, determining a second conditional probability of a second feature occurring in a text given that the text belongs to the classification mode, wherein the second feature is included in the text, determining a probability of the classification mode occurring, multiplying the first conditional probability, the second conditional probability and the probability of the classification mode to determine a product, and storing the product in a tangible memory as a score that the message belongs to the first classification mode.
1. A method to analyze electronic messages, the method comprising:
sorting a set of messages based on a set of probabilities, the set of probabilities indicating a likelihood that the set of messages belong to a first classification;
assigning a first subset of first consecutive ones of the sorted set of messages to a first bucket;
assigning a second subset of second consecutive ones of the sorted set of messages to a second bucket, wherein a first message of the second subset of messages is consecutive to a last message of the first subset of messages with reference to the set of probabilities;
determining a mean of probabilities of the first subset of messages assigned to the first bucket;
storing the mean in a tangible memory; and
storing, for respective ones of the messages assigned to the first bucket, an indication that the respective message is assigned to the first bucket in the tangible memory,
wherein storing the mean in a tangible memory further comprises:
moving a boundary between the first bucket and the second bucket;
recalculating the mean of probabilities of messages assigned to the first bucket in response to moving the boundary;
repeatedly (1) moving the boundary and (2) recalculating the mean until deviations of the probabilities within the first bucket with respect to the mean are equal to or less than a threshold; and
storing a last recalculated mean as representative of the first bucket.
2. The method as defined in claim 1 , wherein the mean is a first mean, and further comprising determining a second mean of messages assigned to the second bucket and storing the second mean in the tangible memory.
3. The method as defined in claim 1 , wherein the mean is stored in the tangible memory in a first data structure associating an identifier for the first bucket with the mean and wherein the set of messages are stored in the tangible memory in a second data structure different from the first data structure, the second data structure associating one of the messages in the set of messages with an identifier for the first bucket.
4. The method as defined in claim 1 , wherein the classification is a sentiment classification.
5. The method as defined in claim 1 , wherein the classification is at least one of a positive classification, a negative classification, a mixed classification or a no opinion classification.
6. The method as defined in claim 1 , wherein the messages are stored in a data structure, further comprising storing, for each one of the messages, an indication of the number of messages that preceded the one of the messages in the data structure.
7. The method according to claim 1 , wherein the indication indicates that the respective message is associated with the mean of probabilities of the first subset of messages.
8. The method according to claim 1 , further comprising retrieving the mean for the first bucket in response to the respective message being included in results of a search query.
9. A method to analyze electronic messages, the method comprising:
sorting a set of messages based on a set of probabilities, the set of probabilities indicating a likelihood that the set of messages belong to a first classification;
assigning a first subset of first consecutive ones of the sorted set of messages to a first bucket;
assigning a second subset of second consecutive ones of the sorted set of messages to a second bucket;
determining a mean of probabilities of the first subset of messages assigned to the first bucket;
storing the mean in a tangible memory;
storing, for each one of the messages assigned to the first bucket, an indication that the respective message is assigned to the first bucket in the tangible memory, wherein the mean is a first mean;
moving a boundary between the first bucket and the second bucket;
in response to moving the boundary, calculating a second mean of the messages assigned to the changed first bucket; and
storing the second mean in the tangible memory when the deviations of the probabilities within the first bucket with respect to the mean are equal to or less than a threshold.
10. A tangible computer readable storage medium comprising instructions that, when executed, cause a machine to at least:
sort a set of messages based on a set of probabilities, the set of probabilities indicating a likelihood that the set of messages belong to a first classification;
assign a first subset of first consecutive ones of the sorted set of messages to a first bucket;
assign a second subset of second consecutive ones of the sorted set of messages to a second bucket, wherein a first message of the second bucket is consecutive to a last message of the first bucket with reference to the set of probabilities;
determine a mean of probabilities of the first subset of messages assigned to the first bucket;
store the mean in a tangible memory;
store, for respective ones of the messages assigned to the first bucket, an indication that the respective message is assigned to the first bucket in the tangible memory,
repeatedly (1) move a boundary between the first bucket and the second bucket and (2) recalculate the mean of probabilities of messages assigned to the first bucket in response to moving the boundary until deviations of the probabilities within the first bucket with respect to the mean are equal to or less than a threshold; and
store a last recalculated mean as representative of the first bucket.
11. The tangible computer readable storage medium as defined in claim 10 , wherein the mean is a first mean, and further comprising determining a second mean of messages assigned to the second bucket and storing the second mean in the tangible memory.
12. The tangible computer readable storage medium as defined in claim 10 , wherein the mean is stored in the tangible memory in a first data structure associating an identifier for the first bucket with the mean and wherein the set of messages are stored in the tangible memory in a second data structure different from the first data structure, the second data structure associating one of the messages in the set of messages with an identifier for the first bucket.
13. The tangible computer readable storage medium as defined in claim 10 , wherein the classification is a sentiment classification.
14. The tangible computer readable storage medium as defined in claim 10 , wherein the classification is at least one of a positive classification, a negative classification, a mixed classification or a no opinion classification.
15. The tangible computer readable storage medium as defined in claim 10 , wherein the messages are stored in a data structure, further comprising storing, for each one of the messages, an indication of the number of messages that preceded the one of the messages in the data structure.
16. A tangible computer readable storage medium comprising instructions that, when executed, cause a machine to at least:
sort a set of messages based on a set of probabilities, the set of probabilities indicating a likelihood that the set of messages belong to a first classification;
assign a first subset of first consecutive ones of the sorted set of messages to a first bucket;
assign a second subset of second consecutive ones of the sorted set of messages to a second bucket;
determine a mean of probabilities of the first subset of messages assigned to the first bucket;
store the mean in a tangible memory;
store, for corresponding ones of the messages assigned to the first bucket, an indication that the respective message is assigned to the first bucket in the tangible memory, wherein the mean is a first mean;
move a boundary between the first bucket and the second bucket;
in response to moving the boundary, calculate a second mean of the messages assigned to the changed first bucket; and
store the second mean in the tangible memory when the deviations of the probabilities within the first bucket with respect to the mean are equal to or less than a threshold.
17. An apparatus to analyze a set of electronic messages, the apparatus comprising:
a statistics generator to generate a set of probabilities that the set of messages belong to a first classification;
a tangible memory; and
an index generator to sort the set of messages based on a set of probabilities, assign a first subset of first consecutive ones of the sorted set of messages to a first bucket, assign a second subset of second consecutive ones of the sorted set of messages to a second bucket, a first message of the second bucket being consecutive to a last message of the first bucket with reference to the set of probabilities, determine a mean of probabilities of the first subset of messages assigned to the first bucket, store the mean in the tangible memory, and store, for each one of the messages assigned to the first bucket, an indication that the respective message is assigned to the first bucket in the tangible memory,
the index generator is further to move a boundary between the first bucket and the second bucket, recalculate the mean of probabilities of messages assigned to the first bucket in response to moving the boundary, repeatedly (1) move the boundary and (2) recalculate the mean until deviations of the probabilities within the first bucket with respect to the mean are equal to or less than a threshold, and store a last recalculated mean as a representative of the first bucket.
18. The apparatus as defined in claim 17 , wherein the mean is a first mean, and the index generator is further to determine a second mean of messages assigned to the second bucket and storing the second mean in the tangible memory.
19. The apparatus as defined in claim 17 , wherein the mean is stored in the tangible memory in a first data structure associating an identifier for the first bucket with the mean and wherein the set of messages are stored in the tangible memory in a second data structure different from the first data structure, the second data structure associating one of the messages in the set of messages with an identifier for the first bucket.
20. The apparatus as defined in claim 17 , wherein the classification is a sentiment classification.
21. The apparatus as defined in claim 17 , wherein the classification is at least one of a positive classification, a negative classification, a mixed classification or a no opinion classification.
22. An apparatus to analyze a set of electronic messages, the apparatus comprising:
a statistics generator to generate a set of probabilities that the set of messages belong to a first classification;
a tangible memory; and
an index generator to sort the set of messages based on a set of probabilities, assign a first subset of first consecutive ones of the sorted set of messages to a second bucket, determine a mean of probabilities of the first subset of messages assigned to the first bucket, store the mean in the tangible memory, and store for each one of the messages assigned to the first bucket, an indication,
wherein the index generator is further to:
move a boundary between the first bucket and the second bucket;
in response to moving the boundary, calculate a second mean of the messages assigned to the changed first bucket; and
store the second mean in the tangible memory when the deviations of the probabilities within the first bucket with respect to the mean are equal to or less than a threshold.