Identifying user intent from social media posts and text data

Abstract

Analyzing the text data and social media posts to obtain an accurate measure of audience interest including business goal features includes collecting text data based on each business goal feature; extracting information from the text data including metadata, actions, and entities along with associated connections; using an intent identifier to identify intents including associated entities based on the extracted information; using the extracted information to filter and recognize relevant input data based on the intent criteria; and providing aggregate data for each business goal feature as feedback for the intent.

Description

[Technical Field]

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS This application claims the benefit of priority under 35 U.S.C. § 119 of co-pending U.S. Provisional Patent Application No. 63 / 105,026, entitled "User Intent Identification from Social Media Post and Text Data," filed October 23, 2020, the disclosure of which is incorporated herein by reference.

[0002] The present disclosure relates to extracting intent from text data, and more particularly to analyzing text data and social media posts to obtain an accurate measure of audience interest by extracting user intent from the text data. [Background technology]

[0003] Current methods for extracting intent from text data are based on sentiment analysis and keyword search. While these provide useful initial clues for any text data, such as social media posts, they are inaccurate due to noise in the text data and too general for deeper business insights. A common goal in marketing applications requires a systematic understanding of audience interests, such as using signals from social media data to predict the likelihood of a commercial surprise or flop. Thus, intent is an action or opinion related to a subject of interest. This subject can be a product, service, or other related topic. Summary of the Invention [Means for solving the problem]

[0004] The present disclosure analyzes text data and social media posts to obtain an accurate measure of audience interest by extracting user intent from the text data and social media posts.

[0005] In one implementation, a system for analyzing text data and social media posts to obtain an accurate measure of audience interest including business target features is disclosed. The system includes a data aggregation unit that collects text data based on at least one of the business target features, an intent identification unit that includes an information extractor and an intent identifier, and a method for measuring accurate audience interest, wherein the information extractor extracts information including metadata, actions, and entities along with associated connections from the collected text data, the information extractor extracts the information using a tool that identifies the role or feature set of each word, and the intent identifier identifies intent actions including associated entities by aggregating general actions toward a goal based on the extracted information.

[0006] In one implementation, the intent identification further includes a classifier trained to assign at least one label, which assigns at least one label to each piece of collected text data, and a scorer that scores each piece of labeled data based on the training and assigns an intent based on the assigned label. In one implementation, the scorer adds a probability to the assigned label, which indicates how likely each piece of labeled data is to belong to the assigned label. In one implementation, the data aggregation is coupled to the classifier and the information extractor so that text data collected from the data aggregation is sent to the classifier and the information extractor in parallel. In one implementation, both the scorer and the intent identifier are coupled to a feedback loop so that outputs from the scorer and the intent identifier are used in a weighted balance. In one implementation, the output of the intent identifier is coupled to an input of the classifier so that extracted information without a clearly identified intent is sent to the classifier. In one implementation, the intent identifier is coupled to a feedback loop so that extracted information with a clearly identified intent is sent to the feedback loop.

[0007] In another implementation, a method is disclosed for analyzing text data and social media posts to obtain an accurate measure of audience interest including business goal features, the method includes: collecting text data based on each business goal feature; extracting information from the text data, including metadata, actions, and entities along with associated connections; identifying intents, including associated entities, based on the extracted information using an intent identifier; filtering and recognizing relevant input data based on the intent criteria using the extracted information; and providing aggregate data for each business goal feature as feedback for the intent.

[0008] In one implementation, the information is extracted using a tool that identifies the role of each word. In one implementation, the intent is identified by aggregating common concepts or actions toward a goal. In one implementation, the method further includes assigning at least one label to each piece of data in the collected text data using a trained classifier. In one implementation, the method further includes using a scorer to score each labeled piece of data based on the training and assigning an intent based on the assigned label. In one implementation, the feedback uses a weighted balance between the output of the intent identifier and the output of the scorer. In one implementation, extracting the information is performed by an information extractor. In one implementation, the method further includes applying the collected text data to both the classifier and the information extractor in parallel. In one implementation, the method further includes sending extracted information with a clearly identified intent to feedback and sending extracted information without a clearly identified intent to the classifier.

[0009] In another implementation, a non-transitory computer-readable storage medium storing a computer program for analyzing text data and social media posts to obtain an accurate measure of audience interest including business goal features is disclosed. The computer program includes executable instructions that cause the computer to collect text data based on each business goal feature, extract information from the text data including metadata, actions, and entities along with associated connections, identify intents including associated entities based on the extracted information using an intent identifier, filter and recognize relevant input data based on the intent criteria using the extracted information, and provide aggregate data related to each business goal feature as feedback related to the intent.

[0010] In one implementation, the computer-readable storage medium further includes executable instructions for causing a computer to assign at least one label to each piece of collected text data. In one implementation, the computer-readable storage medium further includes executable instructions for causing a computer to score each labeled piece of data based on training and assign an intent based on the assigned label. In one implementation, the information is extracted using a tool that identifies the role of each word.

[0011] Other features and advantages will become apparent from the present specification, which illustrates by way of example embodiments of the disclosure.

[0012] Details of the present disclosure, both as to its structure and operation, may be gleaned in part by studying the accompanying drawings, in which like parts are designated with like reference numerals, and in which: [Brief explanation of the drawings]

[0013] [Figure 1A] FIG. 1 is a block diagram of a system that analyzes text data and social media posts to obtain an accurate measure of audience interest, according to one implementation of the present disclosure. [Figure 1B] FIG. 1 is a detailed block diagram of intent identification according to one implementation of the present disclosure. [Figure 1C] FIG. 1 is a block diagram of a system that analyzes text data and social media posts to obtain an accurate measure of audience interest, according to another implementation of the present disclosure. [Figure 1D] FIG. 1 is a block diagram of a system that analyzes text data and social media posts to obtain an accurate measure of audience interest, according to another implementation of the present disclosure. [Figure 2A]This diagram shows an example of processing the tweet "I am going to watch Zombieland soon" and distinguishing between the action "going to watch" and the intent "Zombieland" by "I." [Figure 2B] Figure 1 shows another example of processing the tweet "The city seems like a Zombieland" to identify the action "seems like", the purpose "Zombieland", and the source "the city". [Figure 2C] This is another detailed example of processing the tweet, "I'm nervous to see Bad Boys 3 because I think my favorite has lost its funny and I don't want to face the truth." [Figure 3] FIG. 1 is a flow diagram of a method for analyzing text data and social media posts to obtain an accurate measure of audience interest including business goal characteristics, according to one implementation of the present disclosure. [Figure 4A] 1 is a representation of a computer system and a user according to one implementation of the present disclosure. [Figure 4B] FIG. 1 is a functional block diagram illustrating a computer system hosting a text analytics application according to one implementation of the present disclosure. DETAILED DESCRIPTION OF THE INVENTION

[0014] As mentioned above, current intent extraction from text data is based on sentiment analysis, which can provide an inaccurate measure of audience interest due to noise in the text data. Sentiment analysis trains a classifier to assign a sentiment label (e.g., "positive," "negative," or "neutral") to each collected piece of data, scores each labeled piece of data to indicate the likelihood that the data belongs to that sentiment label, and assigns an intent based on the assigned sentiment label. Thus, a high proportion of data labeled "positive" is considered to reflect a particular action (e.g., going to the movies). Therefore, sentiment analysis often fails to reliably and clearly understand user intent on social media for business purposes for various reasons, such as: (a) it is heavily based on pre-trained data for sentiment analysis; (b) current sentiment tools and methodologies are limited to only a few categories, while intent may include many more types of categories; (c) the same kind of sentiment does not necessarily indicate the same type of intent; and (d) intent identification involves searching for possible future actions from users, as the user's current opinion sentiment may not indicate such intent.

[0015] Some implementations of the present disclosure analyze text data and social media posts to obtain an accurate measure of audience interest by extracting intent from the text data and social media posts. After reading the following description, it will become apparent how to implement the present disclosure in various implementations and applications. While various implementations of the present disclosure are described herein, it should be understood that these implementations are presented by way of example only, and not limitation. Therefore, the detailed description of various implementations should not be construed as limiting the scope or breadth of the present disclosure.

[0016] Features provided in implementations that analyze text data and social media posts to obtain an accurate measure of audience interest may include, but are not limited to, one or more of the following items: (a) data aggregation, (b) information extraction, (c) intent identification, (d) feedback to obtain an accurate measure of audience interest, and (e) definition of new intents or removal / updating of old intents to recognize intent.

[0017] 1A is a block diagram of a system 100 that analyzes text data and social media posts to obtain an accurate measure of audience interest, according to one implementation of the present disclosure. In the example implementation of FIG. 1A, system 100 includes data aggregation 102, intent identification 104, and feedback 106. In one implementation, intent identification 104 includes information extraction.

[0018] In one implementation, data aggregation 102 includes collecting text data based on each business objective feature. For example, tweets about movies can be collected.

[0019] In one implementation, feedback 106 to obtain an accurate measure of audience interest includes providing aggregate data about the target as intent feedback or general opinions. In another implementation, intent categories may change at different analysis stages. For example, initially, "buying tickets" and "watching a movie" may be collected, but later only "watching a movie" may be collected. In a further implementation, feedback is added to collect better data using intents. For example, some movies may be easier to recognize using other words, such as actors. Thus, refinement of data collection can be achieved through iteration as part of feedback on data collection quality.

[0020] 1B is a detailed block diagram of intent identification 104 according to one implementation of the present disclosure. In the example implementation of FIG. 1B, intent identification 104 includes information extractor 110 and intent identifier 112.

[0021] In one implementation, the information extractor 110 extracts metadata, actions, and entities from the text along with associated connections. Additionally, the information extractor 110 extracts information by using tools that identify the role of each word. For example, verb phrases and nouns can be collected from a single tweet.

[0022] In one implementation, the intent identifier 112 identifies intent actions by aggregating general concepts / actions toward a purpose based on extracted information including related entities, and further uses the extracted information to filter and recognize relevant input data based on intent criteria, for example, sampling tweets containing the action of watching a movie.

[0023] 1C is a block diagram of a system 120 that analyzes text data and social media posts to obtain an accurate measure of audience interest, according to another implementation of the present disclosure. In FIG. 1C, system 120 includes data aggregation 102, intent identification 130, and feedback 132. In one implementation, intent identification 130 includes information extraction.

[0024] In one implementation, data aggregation 102 includes collecting text data based on each business objective feature. For example, tweets about movies can be collected.

[0025] In FIG. 1C, the text data collected by data aggregation 102 is applied in parallel, with trained classifier 122 / scorer 124 adding probabilities to labels and information extractor 126 / intent identifier 128 discovering data with clear intent.

[0026] In the exemplary implementation of Figure 1C, system 120, in contrast to system 100 of Figure 1A, involves combining intent identification with training a classifier for supervised labeling. In Figure 1C, intent identification 130 includes classifier 122, scorer 124, information extractor 126, and intent identifier 128.

[0027] In one implementation, the classifier 122 is trained to assign at least one label (e.g., “promotion,” “intent,” “positive,” and “other”) to each piece of data collected by the data aggregation 102. For example, a tweet is assigned as one of the labels defined above (e.g., “promotion,” “intent,” “positive,” or “other”).

[0028] In one implementation, the scorer 124 scores each labeled data based on training and assigns an intent based on the assigned label, such that a high percentage of data labeled "positive" is considered to reflect a particular action (e.g., going to the movies).

[0029] 1C, information extractor 126 extracts metadata, actions, and entities from text along with associated connections. Additionally, information extractor 126 extracts information by using tools that identify the role of each word. For example, verb phrases and nouns can be collected from a single tweet.

[0030] 1C, the intent identifier 128 identifies intent actions based on the extracted information including related entities. The extracted information (extracted by the information extractor 126) is then used to filter and recognize related input data based on intent criteria. For example, tweets containing the action of watching a movie are sampled.

[0031] In the exemplary implementation of FIG. 1C , feedback 132 to obtain an accurate measure of audience interest combines the outputs from the trained classifier 122 / scorer 124 and the information extractor 126 / intent identifier 128. As described above, the trained classifier 122 / scorer 124 combination adds probabilities to labels, and the information extractor 126 / intent identifier 128 combination discovers data with clear intent. In this case, the outputs from the two paths can be combined in a weighted balance according to their contribution to business strategy refinement. For example, text with clear intent can have a higher importance than text identified by the second path.

[0032] 1D is a block diagram of a system 150 that analyzes text data and social media posts to obtain an accurate measure of audience interest, according to another implementation of the present disclosure. In FIG. 1D, system 150 includes data aggregation 102, intent identification 150, and feedback 152. In one implementation, intent identification 150 includes information extraction.

[0033] In one implementation, data aggregation 102 includes collecting text data based on each business objective feature. For example, tweets about movies can be collected.

[0034] 1D , input text data is applied sequentially. For example, input text data collected by data aggregation 102 can first be sent to information extractor 146 and intent identifier 148 to discover data with a clear intent. Input text data for which no clear intent has been identified can then be sent to trained classifier 142 and scorer 144 to add probabilities to the labels.

[0035] In one implementation, the classifier 142 is trained to assign at least one label (e.g., “promotion,” “intent,” “positive,” and “other”) to each piece of data collected by the data aggregation 102. For example, a tweet is assigned as one of the labels defined above (e.g., “promotion,” “intent,” “positive,” or “other”).

[0036] In one implementation, the scorer 144 scores each labeled data based on training and assigns an intent based on the assigned label, such that a high percentage of data labeled "positive" is considered to reflect a particular action (e.g., going to the movies).

[0037] 1D, information extractor 146 extracts metadata, actions, and entities from text along with associated connections. Additionally, information extractor 146 extracts information by using tools that identify the role of each word. For example, verb phrases and nouns can be collected from a single tweet.

[0038] 1D , the intent identifier 148 identifies intent actions based on the extracted information including related entities. The extracted information (extracted by the information extractor 146) is then used to filter and recognize related input data based on intent criteria. For example, tweets containing the action of watching a movie are sampled.

[0039] 1D , input text data is applied sequentially. For example, input text data collected by data aggregation 102 can first be sent to information extractor 146 and intent identifier 148 to discover data 160 with a clear intent. Input text data 162 for which no clear intent was identified is then sent to trained classifier 142 and scorer 144 to add labels with probabilities to the text data in output 164.

[0040] In the exemplary implementation of FIG. 1D , feedback 132 for obtaining an accurate measure of audience interest combines output 160 from the information extractor 146 / intent identifier 148 with output 164 from the trained classifier 142 / scorer 144. As described above, the information extractor 146 / intent identifier 148 combination finds data 160 with a clear intent, and the trained classifier 142 / scorer 144 combination adds a probability-containing label to data without a clearly identified intent to generate output 164. In this case, the outputs 160, 164 from the two paths can be combined in a weighted balance according to their contribution to business strategy refinement. For example, text 160 with a clear intent can be given a higher importance than text 164 identified by the second path.

[0041] One use case aims to identify user intent: "Is the user planning to watch a particular movie?" In this case, the evaluation is based on two metrics: (1) how many of all movies classified by human manual identification as likely to watch the movie were captured as the correct class by the system; and (2) how many of the people identified by the system as likely to watch the movie actually belong to the correct prediction, or the class labeled by the human as likely to watch the movie. Using currently available sentiment analysis, metric (1) received 57.0% and metric (2) received 56.5%. In contrast, using the implementations of Figures 1B, 1C, or 1D described above, metric (1) received 72.3% and metric (2) received 70.6%. Thus, the implementation described above is provided to extract and identify social media user intent for the purpose of reviewing business objectives. The intent is an action or opinion regarding a goal and its related concepts.

[0042] Figure 2A shows an example where a tweet 200, "I am going to watch Zombieland soon," is processed to identify the action "going to watch" and the "I" intent "Zombieland" (see 202). Thus, the intent 204 to watch the movie has been identified along with the corresponding action of watching the movie.

[0043] 2B illustrates another example where a tweet 210, "The city seems like a Zombieland," is processed to identify the action "seems like," the intent "Zombieland," and the source "the city" (see 212). Thus, the intent 214 to watch the desired movie is not identified because the identified action in this tweet 210 is not related to watching the desired movie.

[0044] Figure 2C shows another detailed example of processing a tweet 220, "I'm nervous to see Bad Boys 3 because I think my favorite has lost his funny and I don't want to face the truth." Item 222 shows the extracted information for the process, where the action "see" and the desired movie "Bad Boys 3" are identified. Therefore, the intent 224 to watch the desired movie is identified along with the corresponding action "see the movie (Bad Boy 3)."

[0045] 3 is a flow diagram of a method 300 for analyzing text data and social media posts to obtain an accurate measure of audience interest that includes business goal features, according to one implementation of the present disclosure. In the example implementation of FIG. 3, text data is collected based on each business goal feature at 310. For example, tweets about movies can be collected.

[0046] Next, at 320, information including metadata, actions, and entities is extracted from the text data along with associated connections. In one implementation, the information is extracted by using a tool that identifies the role of each word. For example, verb phrases and nouns can be collected from a single tweet. At 330, intent actions are identified by aggregating general concepts / actions toward goals based on the extracted information including associated entities. Further, at 340, the extracted information is used to filter and recognize relevant input data based on intent criteria. For example, tweets containing the action of watching a movie are sampled. At 350, the aggregated data regarding the goals is provided as feedback or general opinions regarding the intent.

[0047] The advantages of the above-mentioned method include: (a) the method applies to a wide range of categories of user intents; (b) the ability to define intent categories based on a set of actions or a set of entities; (c) the ability to cluster all existing intents; and (d) the ability to reduce potential bias in the training data by making information extraction independent of the type of intent.

[0048] Figure 4A is a representation of a computer system 400 and a user 402 according to an implementation of the present disclosure. As shown and described with respect to systems 100, 120, and 140 of Figures 1A, 1B, and 1C, respectively, and method 300 of Figure 3, user 402 uses computer system 400 to run a text analysis application 490 that reduces the data used during capture.

[0049] The computer system 400 stores and executes the text analysis application 490 of Figure 4B. The computer system 400 can also communicate with a software program 404. The software program 404 can include software code for the text analysis application 490. As described further below, the software program 404 can be loaded onto an external medium such as a CD, DVD, or storage drive.

[0050] Additionally, computer system 400 may be connected to network 480. Network 480 may be connected in a variety of different architectures, such as a client-server architecture, a peer-to-peer network architecture, or other types of architectures. For example, network 480 may communicate with server 485, which coordinates engines and data used in text analysis application 490. Network 480 may also be different types of networks. For example, network 480 may be the Internet, a local area network or any variation of a local area network, a wide area network, a metropolitan area network, an intranet or extranet, or a wireless network.

[0051] 4B is a functional block diagram illustrating a computer system 400 hosting a text analytics application 490 according to an implementation of the present disclosure. The controller 410 is a programmable processor that controls the operation of the computer system 400 and its components. The controller 410 loads instructions (e.g., in the form of a computer program) from memory 420 or an embedded controller memory (not shown) and executes these instructions to control the system, such as for data processing. In its execution, the controller 410 provides a software service to the text analytics application 490. Alternatively, this service may be implemented as a separate hardware component within the controller 410 or the computer system 400.

[0052] Memory 420 temporarily stores data used by other components of computer system 400. In one implementation, memory 420 is implemented as RAM. In one implementation, memory 420 also includes long-term or permanent memory, such as flash memory and / or ROM.

[0053] Storage 430 provides temporary or long-term storage of data used by other components of computer system 400. For example, storage 430 stores data used by text analysis application 490. In one implementation, storage 430 is a hard disk drive.

[0054] Media device 440 accepts removable media and reads and / or writes data to the inserted media. For example, in one implementation, media device 440 is an optical disk drive.

[0055] User interface 450 includes components that accept user input from a user of computer system 400 and present information to user 402. In one implementation, user interface 450 includes a keyboard, a mouse, audio speakers, and a display. Controller 410 uses input from user 402 to coordinate the operation of computer system 400.

[0056] I / O interface 460 includes one or more I / O ports for connecting to corresponding I / O devices, such as external storage devices or auxiliary devices (e.g., printers or PDAs). In one implementation, the ports of I / O interface 460 include ports such as USB ports, PCMCIA ports, serial ports, and / or parallel ports. In another implementation, I / O interface 460 includes a wireless interface for communicating wirelessly with external devices.

[0057] Network interface 470 includes wired and / or wireless network connections, such as an RJ-45 supporting an Ethernet connection or a "Wi-Fi" interface (including but not limited to 802.11).

[0058] Computer system 400 includes additional hardware and software typical of computer systems (e.g., power supplies, cooling, operating systems), but these components are not specifically shown in Figure 4B for simplicity. Other implementations may use different configurations of computer systems (e.g., different bus or storage configurations, or multi-processor configurations).

[0059] In one implementation, each of systems 100, 120, 140 is a system that is entirely comprised of hardware, including one or more digital signal processors (DSPs), general-purpose microprocessors, application specific integrated circuits (ASICs), field programmable gate / logic arrays (FPGAs), or other equivalent integrated or discrete logic circuits. In another implementation, each of systems 100, 120, 140 is comprised of a combination of hardware and software.

[0060] The descriptions of the implementations disclosed herein are provided to enable any person skilled in the art to make or use the invention. Numerous modifications of these implementations will be readily apparent to those skilled in the art, and the principles defined herein may be applied to other implementations without departing from the spirit or scope of the invention. Thus, the present disclosure is not intended to be limited to the implementations shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0061] Those skilled in the art will understand that the various exemplary modules and method steps described herein can be implemented as electronic hardware, software, firmware, or a combination thereof. To clearly illustrate this interchangeability between hardware and software, the various exemplary modules and method steps have been described herein generally in terms of their functionality. Whether such functionality is implemented as hardware or software depends on the particular application and design constraints imposed on the overall system. Those skilled in the art may implement the described functionality in various ways for each particular application, and such implementation decisions should not be interpreted as causing a departure from the scope of the invention. Additionally, the grouping of functions within a module or step is for ease of description. Certain functionality may also be moved from one module or step to another without departing from the disclosure.

[0062] Not all features of each of the above-described embodiments are necessarily required for a particular implementation of the present disclosure. Moreover, it should be understood that the description and drawings set forth herein are representative of the subject matter broadly contemplated by the present invention. Moreover, it should be understood that the scope of the present disclosure fully embraces other implementations that would be apparent to one skilled in the art, and therefore is not to be limited except by the scope of the appended claims. [Explanation of symbols]

[0063] 310 Collect text data based on the characteristics of each business goal 320 Extracting metadata, actions, and entities from text data along with associated connections 330 Identify intent actions based on extracted information (metadata, actions, and entities) 340 Extracted information is used to filter and recognize relevant input data based on intent criteria. 350 Providing aggregate data on goals as feedback or general comments on intent

Claims

1. 1. A system for analyzing text data and social media posts to obtain an accurate measure of audience interest, including business objective characteristics, comprising: data aggregation to collect text data based on at least one of the business objective features; intent identification, including an information extractor and an intent identifier; Equipped with The information extractor extracts information from the collected text data, including metadata, actions, and entities along with associated connections, and the information extractor extracts information using a tool that identifies the role or feature set of each word; The intent identifier identifies intent actions that include related entities by aggregating general actions toward a goal based on the extracted information; and The intent identification is assigning at least one label to each data item in the collected text data; and a classifier trained to assign the at least one label. a scorer that scores each labeled data based on training and assigns an intent based on the assigned label; further comprising an output of the intent identifier is coupled to an input of the classifier such that the extracted information that does not have a clearly identified intent is sent to the classifier; A system characterized by:

2. The scorer adds a probability to the assigned label, the probability indicating how likely each labeled data item is to belong to the assigned label. The system of claim 1 .

3. the data aggregation is coupled to the classifier and the information extractor such that the text data collected from the data aggregation is sent to the classifier and the information extractor in parallel. The system of claim 1 .

4. Both the scorer and the intent classifier are coupled in a feedback loop such that the output from the scorer and the output from the intent classifier are used in a weighted balance. The system of claim 1 .

5. the intent identifier combines the extracted information with the feedback so that the extracted information having a clearly identified intent is transmitted to the feedback. The system of claim 4.

6. 1. A method for analyzing text data and social media posts to obtain an accurate measure of audience interest including business objective characteristics, the method comprising: collecting the text data based on each business objective feature; extracting information from the text data, including metadata, actions, and entities along with associated connections; using an intent identifier to identify an intent that includes an associated entity based on the extracted information; using the extracted information to filter and recognize relevant input data based on intent criteria; providing aggregate data regarding each business objective characteristic as feedback regarding the intent; The method includes: assigning at least one label to each data item in the collected text data using a trained classifier; sending the extracted information with a clearly identified intent to the feedback; and sending the extracted information that does not have a clearly identified intent to the classifier; The method further comprising:

7. The information is extracted using a tool that identifies the role of each word. The method of claim 6.

8. Intents are identified by aggregating general concepts or actions toward a goal; The method of claim 6.

9. using a scorer to score each labeled data based on the training and assign an intent based on the assigned label. The method of claim 6.

10. the feedback uses a weighted balance between the output of the intent classifier and the output of the scorer.

10. The method of claim 9.

11. Extracting the information is performed by an information extractor; The method of claim 6.

12. applying the collected text data to both the classifier and the information extractor in parallel. The method of claim 11.

13. 1. A non-transitory computer-readable storage medium storing a computer program for analyzing text data and social media posts to obtain an accurate measure of audience interest including business objective characteristics, the computer program comprising: collecting the text data based on each business objective feature; extracting information from the text data, including metadata, actions, and entities along with associated connections; using an intent identifier to identify an intent that includes an associated entity based on the extracted information; using the extracted information to filter and recognize relevant input data based on intent criteria; providing aggregate data regarding each business goal characteristic as feedback regarding the intent; the computer program comprising executable instructions for causing a computer to: assigning at least one label to each data item in the collected text data using a trained classifier; sending the extracted information with a clearly identified intent to the feedback; and sending the extracted information that does not have a clearly identified intent to the classifier; 10. A computer-readable storage medium further comprising executable instructions for causing the computer to:

14. further comprising executable instructions for causing the computer to assign at least one label to each data item in the collected text data. The computer-readable storage medium of claim 13.

15. further comprising executable instructions for causing the computer to score each labeled data based on training and assign an intent based on the assigned label.

15. The computer-readable storage medium of claim 14.

16. The information is extracted using a tool that identifies the role of each word. The computer-readable storage medium of claim 13.

Images