Online interaction processing

Abstract

Methods and apparatus to process online interactions from numerous sources, coming from different time frames and systems. One embodiment includes a facilitator, an interaction processor, and an action generator. The facilitator facilitates interactions in one or more devices, where interactions typically occur. The processor can consolidate and analyze interactions from multiple sources, in different formats, and collected at different time frames, to extract intelligence from them. The processor processes such interactions, and generates an interaction descriptor for each interaction. A descriptor can include a generalization of the corresponding interaction, and at least a part of the interaction. descriptors for different types of interactions can be represented by the same format to allow the processor to analyze them together. The action generator can improve on future interactions; refer an interaction to be responded by a human representative if necessary; and allow users to extract information from the analysis, and generate reports regarding the interactions. Future interactions can then be significantly enhanced, with customers having higher satisfaction level, and companies more accurately charting their future.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS

This application
Also, in one embodiment, a descriptor has an attribute related a generalization of the corresponding interaction, and has a value for one its attributes.

The descriptor analyzer can analyze the descriptors, or the values associated with attributes in the descriptors, to have an understanding of one interaction, or a number of interactions in aggregate. FIG. 10 shows examples of results from analyzing the descriptors, 600. Based on the analysis, the action generator can generate actionable items, 602. The action generator can also generate rules or combinatorial rules automatically, 604. In another example, the generator can generate rules, which lead to actionable items.

Rules can be generated from features in the descriptors. As an example, a company has an east cost distributor and a west coast distributor. During the previous three months,

• • I. From the east coast customers, one set of descriptors indicate that more than 90% of the interactions are on the newly introduced camping gear model ABC. More than 80% of the interactions are through emails or phone calls to call centers. More than 70% of the emotional state of the interactions is frustrated. Such information can be collected based on analyzing values in different attributes of the descriptors. For example, one subject matter attribute designates camping gear model ABC. To get the 90% number, the analyzer can divide the total number of interactions with that subject matter by the total number of interactions during the previous three months.
  • II. From the west coast customers, another set of descriptors indicates again that more than 90% of the interactions are on the newly introduced camping gear model ABC. More than 75% of the interactions are through self-help query boxes on their Web sites. More than 80% of the interactions do not clearly show their emotional state.
    The analyzer analyzes the two sets of descriptors together because they are under the same subject matter. The analyzer has one prior rule:

If an interaction does not show frustration, the corresponding user is not frustrated. Based on this rule, more than 80% of the west coast users are not frustrated. The fact pattern becomes:

• • >70% of a group of users are frustrated, and 80% of them use email or phone calls.
  • >80% of a group of users are not frustrated, and 75% of them use self-help query box.
    The analyzer can then analyze the fact patterns and create the following rules:
  • Frustrated customers are more inclined to use emails, or phone calls to call centers.
  • Non-frustrated customers are more inclined to use self-help query boxes on Web sites.

In another example, Jane asked a question on credit card payment terms through the natural language search system in the self-help query box; and after some time, she asked a similar question through email. By analyzing both descriptors together, the analyzer can draw the following conclusion:

• • The natural language search system in the self-help query box needs improvement in responding to questions regarding credit card payment terms.

Generating such rules based on the descriptor information should be known to those skilled in the art, and will not be further described.

As another example, based on analyzing a number of descriptors, one can make specific business decisions. The following are observations from descriptors:

• • (a) Users who have spent more than 30 minutes learning about the LawnMower Model Z81 do not buy that model. In this example, descriptors have attributes on duration of interaction, and subject matter on products and whether customers buy the products. So, the above observations can be made.
  • (b) More than 90% of users who spent more than 30 minutes learning about a lawnmower other than the LawnMower Model Z81 buy that model.
  • (c) More than 90% of users who have bought LawnMower Model Z81 spends more than two hours on support.
  • (d) Less than 2% of users who have bought lawnmowers other than LawnMower Model Z81 spends more than two hours on support.

Based on the above fact patterns extracted from the descriptors, the analyzer can draw the following conclusions:

• • From (a) and (b), the more one knows about LawnMower Model Z81, the less chance of him to buy it. From (c) and (d), LawnMower Model Z81 needs more support than other lawnmowers. Managers in the company may then decide that LawnMower Model Z81 is defective and may stop carrying it.

An administrator for a corporation can also add different rules into the action generator. For example, the corporation just introduces a new type of grass cutter. The corporation can add in the rule that any customer asking for lawn mower in the next 60 days would also get an advertisement of that grass cutter. Such rules can then be passed onto one or more facilitators through the interaction enhancer.

The interaction enhancer can generate an actionable item for improving or modifying future interactions. This can be through parametric or non-parametric adjustment.

An example of parametric adjustments is that before analyzing the descriptors, when consumers ask for product information, a company through a facilitator, provides them with information on Toyota Camry first and then Toyota Corolla. However, analyses from many descriptors indicate that more than 50% of consumers are looking for Corolla, and only 10% for Camry. In the future, when consumers ask for product information, the facilitator can adjust so that information on Corolla is displayed before Camry. This can be done automatically, such as having the interaction enhancer automatically changing weights for a key word search engine in the facilitator. Similarly, one can change weights in a natural-language response system so that responses for Toyota Corolla are of higher priority than responses for Toyota Camry. Results on Corolla can then be presented before those related to Camry. In another example, the change can be done manually by an administrator.

An example of non-parametric adjustment is through adding content to the knowledge base of natural language search engines in facilitators. For example, if more than 50% of users asking about fax machines do not select any of the responses, the facilitators should modify the content in the knowledge bases regarding fax machines.

Another type of actionable items can be in generating reports on the interactions. In one embodiment, a user can enter requests into the report generator to get different types of reports regarding the interactions. This can be done through a parametric search engine where the user can enter into the system different parameters to get different reports. To customize reports, one can add new fields. For example, one can add a field in the report that matches and tracks all user profiles and their cellular numbers. Referring back to the previous example on Camry, if the report to the administrator indicates that Camry is not that popular, he can enter a rule into the report generator. As explained above, the rule can modify future interactions, such as through the interaction enhancer, to de-emphasize Camry in future responses.

The reports can be in standard relational database format, or the reports can provide a three dimensional views of the data. In one embodiment, the report generator can be an off-the-shelf product, coupled to the analyzer to produce reports.

A third type of actionable items can be in changing the mode of communication in view of the analysis. For example, a descriptor indicates that a user is quite frustrated. The company might want to escalate the interaction to a human representative to call that specific user on issues described in the interaction. The human representative can be a service representative or an expert in the area of interest of the user. In another approach, the escalator can send a trouble-ticket to a call center. This can then lead to a service representative contacting the user through voice over Internet Protocol, instant messaging, chat in a Web collaboration environment, or just through the telephone. To improve customer satisfaction, before the human representative contacts the user, the representative can receive all prior communications with the user in the last two weeks, and other personal information related to the user that would be helpful to the representative. At least, the user does not have to re-convey all of his prior messages to the representative again. In one approach, this escalation is performed through XML. The escalator can consolidate prior interactions into an XML document, and select another mode of communication for the user. Then, the escalator guides the user to the other mode of communication, along with the XML document. In another embodiment, the escalator can direct the user to a third parties' Web site if it is more appropriately for the third party to resolve the issue.

The response to an escalation can be sent at a later time and through different means. For example, the human representative is activated to call the customer regarding his question. He cannot locate him, and leaves a message. The customer does not call back. Later, a modified answer to his question is generated electronically. This modified answer can be in view of the customer's frustration, or can be in view of frustrations as shown by a number of interactions with similar the similar question. The escalator keeps track of that customer not calling back. When the modified answer becomes available electronically, the escalator can automatically send an email to the customer, asking him if he wants a more appropriate answer to his question through email. If his response is yes, the escalator can automatically send him the modified answer.

In one embodiment, the analyzer can also analyze values associated with the mode of communication entry. These can be specific entries or new entries for data not normally categorized in the existing entries of the descriptors. These specific or new entries can be set by an administrator.

One embodiment includes a security module. This module can be for user level security. It controls the identity or the type of administrators or users that can access and/or update different sets of data. The module can also provide a higher level security, such as controlling the one or more users authorized to change the identity or the type of users that can access and/or update data. In another example, the module is for system level security. It can control the one or more users who can change the configurations of the systems, such as the operating parameters of the report generator or the descriptor generator.

As described, interactions can be based on sound, with voice recognition techniques converting the sound into representations to be analyzed. Interactions can be based on images, with pattern recognition techniques again converting them into representations to be analyzed. In another embodiment, the present invention is also applicable to interactions based on smell, tactile or taste. Similarly, those interactions are converted into representations that can be more efficiently analyzed. For example, pressure sensors can be used to digitize tactile interactions to be analyzed.

One embodiment of the invention is implemented as a Web service by an application service provider. For example, a facilitator, administered by a company, facilitates and stores interactions with their customers, employees and partners. The interactions can be represented in XML format, and transported between the provider and the company in SOAP protocol. An interaction processor and an action generator reside in an application service provider. Through the Internet, the company sends the interactions to the provider, or the provider may just access the interactions from the company's storage media. After processing and again through the Internet, the provider sends actionable items with analysis results to the company, or the company may just access the analysis results from the provider.

In yet another embodiment, all of its components are localized. For example, the embodiment is implemented through software. The source code is separated into two sections. One is related to specific languages, and the other is language independent. To localize the source code for a different language, one only needs to modify the section related to languages.

In one embodiment, components can be written in Java, with the data representation in XML. Rules can be in Java objects, and interfaces among components can be in XML format.

Based on the embodiments, corporations will have a better understanding of their customers, and will have significantly better and more consistent systems to interact with their customers. Sales, service and marketing functions will be able to better work together in presenting a single face to customers through different touch points or devices, across a corporation's relationship network.

Many of the embodiments use customers as examples. However, the present invention is also applicable to employees, vendors and partners. Based on the present invention, corporations, partners and vendors would be able to better work together in multi-parties, many-to-many interactions.

Other embodiments of the invention will be apparent to those skilled in the art from a consideration of this specification or practice of the invention disclosed herein. It is intended that the specification and examples be considered as exemplary only, with the true scope and spirit of the invention being indicated by the following claims.

Claims

1. A computer-implemented method for processing online interactions, said method comprising:

accessing two online interactions;

generating, using a processing unit, two interaction descriptors from the two online interactions using grammatical and semantic processing; and

analyzing, using the processing unit, the two descriptors together to determine a piece of information related to the interactions;

wherein the two interactions are from different time frames,

wherein one descriptor includes a generalization of the corresponding interaction, and

wherein the two interactions are from different modes of communication.

2. A computer-implemented method as recited in claim 1 wherein one interaction is an instant message, and the other interaction is an email.

3. A computer-implemented method as recited in claim 1 wherein one interaction is from a mobile device and the other interaction is from a desktop device.

4. A computer-implemented method as recited in claim 1 wherein the two interactions are from two physical locations.

5. A computer-implemented method as recited in claim 1 wherein the two interactions have two different transport protocols.

6. A computer-implemented method as recited in claim 1 wherein the two interactions have two different data formats.

7. A computer-implemented method as recited in claim 1 wherein the two interactions are in two different languages, and the method further comprises normalizing the two interactions into one language.

8. A computer-implemented method as recited in claim 1 wherein at least one of the interactions includes a pre-determined representation, and the method further comprises normalizing the pre-determined representation into a standard representation.

9. A computer-implemented method as recited in claim 1 wherein one descriptor includes at least a part of the corresponding interaction.

10. A computer-implemented method as recited in claim 1 wherein one descriptor includes an identifier for identifying the corresponding interaction.

11. A computer-implemented method as recited in claim 1 wherein a corporation pays for the computer-implemented method, wherein the generalization is related to a subject matter of the corresponding interaction, and wherein the subject matter is related to an interest of the corporation.

12. A computer-implemented method as recited in claim 1 wherein a person initiates one of the interactions, the generalization is related to an interest of the person, and the interest is extracted from the interaction.

13. A computer-implemented method as recited in claim 1 wherein a person initiates one of the interactions, and the generalization is related to the state of the person.

14. A computer-implemented method as recited in claim 1 further comprising facilitating the interactions before accessing the interactions.

15. A computer-implemented method as recited in claim 1 wherein the piece of information includes a rule generated based on the two interactions.

16. A computer-implemented method as recited in claim 1 wherein the piece of information enhances future interactions.

17. A computer-implemented method as recited in claim 1 wherein the piece of information is for generating a report related to the interactions.

18. A computer-implemented method as recited in claim 1 wherein the piece of information is for changing the mode of communication of one of the interactions.

19. A computer-implemented method as recited in claim 1 wherein future interactions can be modified based on an input from a user who has accessed the piece of information.

20. A computer-implemented method as recited in claim 1 wherein one interaction includes a statement and a response to the statement, wherein based on the piece of information, a human representative is activated to call the initiator of that interaction, wherein based on the piece of information, the response to the statement is modified, and wherein the modified response is sent to the initiator.

21. A method for processing online interactions, the method comprising:

receiving first information regarding a first online interaction, wherein the first information corresponds to a first time frame;

receiving second information regarding a second online interaction, wherein the second information corresponds to a second time frame;

generating, using a processing unit, a first interaction descriptor corresponding to the first online interaction;

generating, using the processing unit, a second interaction descriptor corresponding to the second online interaction;

analyzing, using the processing unit, the first interaction descriptor and the second interaction descriptor; and

identifying, using the processing unit, third information based on the analysis, wherein the third information is related to the first online interaction and the second online interaction.

22. The method of claim 21, further comprising providing a response to the second online interaction, wherein the response is based at least in part on the third information.

23. The method of claim 21, further comprising providing a response to a third online interaction, wherein the response is based at least in part on the third information.

24. The method of claim 21, wherein the first interaction descriptor comprises a generalization of the first online interaction.

25. The method of claim 21, wherein the first information is received according to a first transport protocol and the second information is received according to a second transport protocol.

26. The method of claim 21, wherein the first information is received by way of a first mode of communication and the second information is received by way of a second mode of communication.

27. The method of claim 21, further comprising processing, using the processing unit, the first information to identify grammatical information, wherein the first interaction descriptor is generated based at least in part on the grammatical information.

28. The method of claim 27, further comprising processing, using the processing unit, the first information to identify semantic information, wherein the first interaction descriptor is generated based at least in part on the semantic information.

29. A system for processing online interactions, the system comprising:

a processing unit; and

a facilitator configured to:

receive first information regarding a first online interaction, wherein the first information is received during a first time frame; and

receive second information regarding a second online interaction, wherein the second information is received during a second time frame;

wherein the processing unit is coupled to the facilitator, wherein the processing unit is configured to:

generate a first interaction descriptor corresponding to the first online interaction;

generate a second interaction descriptor corresponding to the second online interaction;

analyze the first interaction descriptor and the second interaction descriptor; and

identify third information based on the analysis, wherein the third information is related to the first online interaction and the second online interaction.

30. The system of claim 29, wherein the first information is in a first language and the second information is in a second language, and further wherein the facilitator is configured to normalize the first information and the second information into a common language.

31. The system of claim 29, wherein the first information comprises a query, and further wherein the facilitator is configured to generate a response to the query, wherein the response is based at least in part on the third information.

32. The system of claim 29, wherein the first information comprises at least one of visual data, voice data, handwriting, or key strokes, and further comprising a pattern recognition system configured to identify a pattern in at least one of the visual data, the voice data, the handwriting, or the key strokes.

33. The system of claim 29, further comprising a report generator configured to generate a report based at least in part on the third information.

34. The system of claim 29, further comprising an escalator configured to escalate the first online transaction to a customer service representative based at least in part on the first interaction descriptor.

35. The system of claim 29, further comprising an interaction enhancer configured to use the third information to enhance a third online interaction.

36. A non-transitory computer-readable storage medium having instructions stored thereon for processing online interactions, wherein upon execution by a processor, the instructions cause the processor to:

access first information regarding a first online interaction, wherein the first information corresponds to a first time frame;

access second information regarding a second online interaction, wherein the second information corresponds to a second time frame;

generate a first interaction descriptor corresponding to the first online interaction;

generate a second interaction descriptor corresponding to the second online interaction;

analyze the first interaction descriptor and the second interaction descriptor; and

identify third information based on the analysis, wherein the third information is related to the first online interaction and the second online interaction.

37. The computer-readable medium of claim 36, wherein the first online interaction comprises a request from a user, and further wherein the first information is related to a state of the user.

38. The computer-readable medium of claim 36, wherein the processor is further caused to generate a rule based at least in part on the third information.