Methods and apparatus for person-centric multichannel opinion mining in data lakes

A computer program product, comprising a tangible machine-readable storage medium having encoded therein executable code of one or more software programs, wherein the one or more software programs when executed by at least one processing device perform the following steps:

• obtaining multi-channel heterogeneous data from a plurality of channels, wherein said plurality of channels comprise at least two of social media channels, call center channels, chat channels, email channels and an Internet channel;
• aggregating said multi-channel heterogeneous data from said plurality of channels into a data lake;
• identifying one or more target entities that are targets of opinion information across said plurality of channels;
• extracting a plurality of user identities from said plurality of channels;
• aligning, using at least one processing device of said data lake, said plurality of extracted user identities across said plurality of channels to link common user identities;
• identifying, using the at least one processing device of said data lake, one or more of said target entities that are targets of said opinion information of said extracted user identities;
• linking, using the at least one processing device of said data lake, said opinion information of said extracted user identities with a linked common user identity associated with an opinion holder that expressed said opinion information;
• determining, using the at least one processing device of said data lake, a sentiment classification of said opinion information indicating whether said opinion information of said opinion holder comprises one or more of a positive opinion and a negative opinion by identifying subjective information within said opinion information and applying a trained classifier to said subjective information to determine said sentiment classification, wherein said trained classifier is trained by applying a supervised learning technique to a training dataset having labeled sentiment classifications;
• clustering, using the at least one processing device of said data lake, a plurality of linked common user identities into a plurality of node clusters such that each node cluster comprises a plurality of linked common user identities expressing opinion information about related target entities;
• providing, using the at least one processing device of said data lake, a summary of a plurality of said opinion information of at least one opinion holder having an associated linked common user identity across said plurality of channels with respect to said one or more target entities;
• and
• initiating, using the at least one processing device of said data lake, a communication to at least one of said linked common user identities in a given node cluster based at least in part on an evaluation of one or more of a frequency of said opinion information and said sentiment classification of said opinion information.

A system, comprising:

• a memory;
• and
• at least one processing device, coupled to the memory, operative to implement the following steps:
• obtaining multi-channel heterogeneous data from a plurality of channels, wherein said plurality of channels comprise at least two of social media channels, call center channels, chat channels, email channels and an Internet channel;
• aggregating said multi-channel heterogeneous data from said plurality of channels into a data lake;
• identifying one or more target entities that are targets of opinion information across said plurality of channels;
• extracting a plurality of user identities from said plurality of channels;
• aligning, using at least one processing device of said data lake, said plurality of extracted user identities across said plurality of channels to link common user identities;
• identifying, using the at least one processing device of said data lake, one or more of said target entities that are targets of said opinion information of said extracted user identities;
• linking, using the at least one processing device of said data lake, said opinion information of said extracted user identities with a linked common user identity associated with an opinion holder that expressed said opinion information;
• determining, using the at least one processing device of said data lake, a sentiment classification of said opinion information indicating whether said opinion information of said opinion holder comprises one or more of a positive opinion and a negative opinion by identifying subjective information within said opinion information and applying a trained classifier to said subjective information to determine said sentiment classification, wherein said trained classifier is trained by applying a supervised learning technique to a training dataset having labeled sentiment classifications;
• clustering, using the at least one processing device of said data lake, a plurality of linked common user identities into a plurality of node clusters such that each node cluster comprises a plurality of linked common user identities expressing opinion information about related target entities;
• providing, using the at least one processing device of said data lake, a summary of a plurality of said opinion information of at least one opinion holder having an associated linked common user identity across said plurality of channels with respect to said one or more target entities;
• and
• initiating, using the at least one processing device of said data lake, a communication to at least one of said linked common user identities in a given node cluster based at least in part on an evaluation of one or more of a frequency of said opinion information and said sentiment classification of said opinion information.